sdl video push conversion stream and toxav video sink

input with reframer (magic fix, someone pls tell my why)

.github/workflows/cd.yml

@@ -22,10 +22,10 @@ jobs:
         submodules: recursive
 
     - name: Install Dependencies
-      run: sudo apt update && sudo apt -y install libsodium-dev cmake
+      run: sudo apt update && sudo apt -y install libsodium-dev cmake libvpx-dev libopus-dev
 
     - name: Configure CMake
-      run: cmake -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}}
+      run: cmake -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DTOMATO_TOX_AV=ON
 
     - name: Build
       run: cmake --build ${{github.workspace}}/build --config ${{env.BUILD_TYPE}} -j 4 -t tomato
@@ -101,7 +101,7 @@ jobs:
     - name: Configure CMake
       env:
         ANDROID_NDK_HOME: ${{steps.setup_ndk.outputs.ndk-path}}
-      run: cmake -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=${{matrix.platform.vcpkg_toolkit}} -DANDROID=1 -DANDROID_PLATFORM=23 -DANDROID_ABI=${{matrix.platform.ndk_abi}} -DVCPKG_CHAINLOAD_TOOLCHAIN_FILE=${{steps.setup_ndk.outputs.ndk-path}}/build/cmake/android.toolchain.cmake -DSDLIMAGE_JPG_SHARED=OFF -DSDLIMAGE_PNG_SHARED=OFF -DTOMATO_MAIN_SO=ON
+      run: cmake -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=${{matrix.platform.vcpkg_toolkit}} -DANDROID=1 -DANDROID_PLATFORM=23 -DANDROID_ABI=${{matrix.platform.ndk_abi}} -DVCPKG_CHAINLOAD_TOOLCHAIN_FILE=${{steps.setup_ndk.outputs.ndk-path}}/build/cmake/android.toolchain.cmake -DSDLIMAGE_JPG_SHARED=OFF -DSDLIMAGE_PNG_SHARED=OFF -DTOMATO_MAIN_SO=ON -DTOMATO_TOX_AV=ON
 
     - name: Build (tomato)
       run: cmake --build ${{github.workspace}}/build --config ${{env.BUILD_TYPE}} -j 4 -t tomato
@@ -164,7 +164,7 @@ jobs:
     #- uses: ilammy/setup-nasm@v1
 
     - name: Configure CMake
-      run: cmake -G Ninja -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_TOOLCHAIN_FILE=C:/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=x64-windows-static -DSDLIMAGE_VENDORED=ON -DSDLIMAGE_DEPS_SHARED=ON -DSDLIMAGE_JXL=OFF -DSDLIMAGE_AVIF=OFF -DPKG_CONFIG_EXECUTABLE=C:/vcpkg/installed/x64-windows/tools/pkgconf/pkgconf.exe
+      run: cmake -G Ninja -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_TOOLCHAIN_FILE=C:/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=x64-windows-static -DSDLIMAGE_VENDORED=ON -DSDLIMAGE_DEPS_SHARED=ON -DSDLIMAGE_JXL=OFF -DSDLIMAGE_AVIF=OFF -DPKG_CONFIG_EXECUTABLE=C:/vcpkg/installed/x64-windows/tools/pkgconf/pkgconf.exe -DTOMATO_TOX_AV=ON
 
     - name: Build
       run: cmake --build ${{github.workspace}}/build --config ${{env.BUILD_TYPE}} -t tomato
@@ -229,7 +229,7 @@ jobs:
     #- uses: ilammy/setup-nasm@v1
 
     - name: Configure CMake
-      run: cmake -G Ninja -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_TOOLCHAIN_FILE=C:/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=x64-windows-static -DTOMATO_ASAN=ON -DCMAKE_MSVC_RUNTIME_LIBRARY=MultiThreaded -DSDLIMAGE_VENDORED=ON -DSDLIMAGE_DEPS_SHARED=ON -DSDLIMAGE_JXL=OFF -DSDLIMAGE_AVIF=OFF -DPKG_CONFIG_EXECUTABLE=C:/vcpkg/installed/x64-windows/tools/pkgconf/pkgconf.exe
+      run: cmake -G Ninja -B ${{github.workspace}}/build -DCMAKE_BUILD_TYPE=${{env.BUILD_TYPE}} -DCMAKE_TOOLCHAIN_FILE=C:/vcpkg/scripts/buildsystems/vcpkg.cmake -DVCPKG_TARGET_TRIPLET=x64-windows-static -DTOMATO_ASAN=ON -DCMAKE_MSVC_RUNTIME_LIBRARY=MultiThreaded -DSDLIMAGE_VENDORED=ON -DSDLIMAGE_DEPS_SHARED=ON -DSDLIMAGE_JXL=OFF -DSDLIMAGE_AVIF=OFF -DPKG_CONFIG_EXECUTABLE=C:/vcpkg/installed/x64-windows/tools/pkgconf/pkgconf.exe -DTOMATO_TOX_AV=ON
 
     - name: Build
       run: cmake --build ${{github.workspace}}/build --config ${{env.BUILD_TYPE}} -j 4 -t tomato

flake.nix

@@ -80,6 +80,7 @@
         ] ++ self.packages.${system}.default.dlopenBuildInputs;
 
         cmakeFlags = [
+          "-DTOMATO_TOX_AV=ON"
           "-DTOMATO_ASAN=OFF"
           "-DCMAKE_BUILD_TYPE=RelWithDebInfo"
           #"-DCMAKE_BUILD_TYPE=Debug"

src/CMakeLists.txt

@@ -102,12 +102,36 @@ target_sources(tomato PUBLIC
 
 	./chat_gui4.hpp
 	./chat_gui4.cpp
+
+	./frame_streams/frame_stream2.hpp
+	./frame_streams/audio_stream2.hpp
+	./frame_streams/stream_manager.hpp
+	./frame_streams/stream_manager.cpp
+	./frame_streams/locked_frame_stream.hpp
+	./frame_streams/multi_source.hpp
+
+	./frame_streams/voip_model.hpp
+
+	./frame_streams/sdl/sdl_audio2_frame_stream2.hpp
+	./frame_streams/sdl/sdl_audio2_frame_stream2.cpp
+	./frame_streams/sdl/video.hpp
+	./frame_streams/sdl/video_push_converter.hpp
+	./frame_streams/sdl/video_push_converter.cpp
+
+	./stream_manager_ui.hpp
+	./stream_manager_ui.cpp
+
+	./debug_video_tap.hpp
+	./debug_video_tap.cpp
 )
 
 if (TOMATO_TOX_AV)
 	target_sources(tomato PUBLIC
 		./tox_av.hpp
 		./tox_av.cpp
+
+		./tox_av_voip_model.hpp
+		./tox_av_voip_model.cpp
 	)
 
 	target_compile_definitions(tomato PUBLIC TOMATO_TOX_AV)
@@ -147,3 +171,18 @@ target_link_libraries(tomato PUBLIC
 
 set_target_properties(tomato PROPERTIES POSITION_INDEPENDENT_CODE ON)
 
+########################################
+
+add_executable(test_frame_stream2_pop_reframer EXCLUDE_FROM_ALL
+	./frame_streams/frame_stream2.hpp
+	./frame_streams/audio_stream2.hpp
+	./frame_streams/locked_frame_stream.hpp
+	./frame_streams/multi_source.hpp
+
+	./frame_streams/test_pop_reframer.cpp
+)
+
+target_link_libraries(test_frame_stream2_pop_reframer
+	solanaceae_util
+)
+

src/chat_gui/file_selector.cpp

@@ -111,56 +111,60 @@ void FileSelector::render(void) {
 				}
 			}
 
-			// do sorting here
-			// TODO: cache the result (lol)
-			if (ImGuiTableSortSpecs* sorts_specs = ImGui::TableGetSortSpecs(); sorts_specs != nullptr && sorts_specs->SpecsCount >= 1) {
-				switch (static_cast<SortID>(sorts_specs->Specs->ColumnUserID)) {
-					break; case SortID::name:
-						if (sorts_specs->Specs->SortDirection == ImGuiSortDirection_Descending) {
-							std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
-								return a.path() < b.path();
-							});
-							std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
-								return a.path().filename() < b.path().filename();
-							});
-						} else {
-							std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
-								return a.path() > b.path();
-							});
-							std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
-								return a.path().filename() > b.path().filename();
-							});
-						}
-					break; case SortID::size:
-						if (sorts_specs->Specs->SortDirection == ImGuiSortDirection_Descending) {
-							// TODO: sort dirs?
-							std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
-								return a.file_size() < b.file_size();
-							});
-						} else {
-							// TODO: sort dirs?
-							std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
-								return a.file_size() > b.file_size();
-							});
-						}
-					break; case SortID::date:
-						if (sorts_specs->Specs->SortDirection == ImGuiSortDirection_Descending) {
-							std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
-								return a.last_write_time() < b.last_write_time();
-							});
-							std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
-								return a.last_write_time() < b.last_write_time();
-							});
-						} else {
-							std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
-								return a.last_write_time() > b.last_write_time();
-							});
-							std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
-								return a.last_write_time() > b.last_write_time();
-							});
-						}
-					break; default: ;
+			try {
+				// do sorting here
+				// TODO: cache the result (lol)
+				if (ImGuiTableSortSpecs* sorts_specs = ImGui::TableGetSortSpecs(); sorts_specs != nullptr && sorts_specs->SpecsCount >= 1) {
+					switch (static_cast<SortID>(sorts_specs->Specs->ColumnUserID)) {
+						break; case SortID::name:
+							if (sorts_specs->Specs->SortDirection == ImGuiSortDirection_Descending) {
+								std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
+									return a.path() < b.path();
+								});
+								std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
+									return a.path().filename() < b.path().filename();
+								});
+							} else {
+								std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
+									return a.path() > b.path();
+								});
+								std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
+									return a.path().filename() > b.path().filename();
+								});
+							}
+						break; case SortID::size:
+							if (sorts_specs->Specs->SortDirection == ImGuiSortDirection_Descending) {
+								// TODO: sort dirs?
+								std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
+									return a.file_size() < b.file_size();
+								});
+							} else {
+								// TODO: sort dirs?
+								std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
+									return a.file_size() > b.file_size();
+								});
+							}
+						break; case SortID::date:
+							if (sorts_specs->Specs->SortDirection == ImGuiSortDirection_Descending) {
+								std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
+									return a.last_write_time() < b.last_write_time();
+								});
+								std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
+									return a.last_write_time() < b.last_write_time();
+								});
+							} else {
+								std::sort(dirs.begin(), dirs.end(), [](const auto& a, const auto& b) -> bool {
+									return a.last_write_time() > b.last_write_time();
+								});
+								std::sort(files.begin(), files.end(), [](const auto& a, const auto& b) -> bool {
+									return a.last_write_time() > b.last_write_time();
+								});
+							}
+						break; default: ;
+					}
 				}
+			} catch (...) {
+				// we likely saw a file disapear
 			}
 
 			for (auto const& dir_entry : dirs) {

src/chat_gui4.cpp

@@ -9,9 +9,13 @@
 #include <solanaceae/contact/components.hpp>
 #include <solanaceae/util/utils.hpp>
 
+#include "./frame_streams/voip_model.hpp"
+
 // HACK: remove them
 #include <solanaceae/tox_contacts/components.hpp>
 
+#include <entt/entity/entity.hpp>
+
 #include <imgui/imgui.h>
 #include <imgui/misc/cpp/imgui_stdlib.h>
 
@@ -30,6 +34,7 @@
 #include <fstream>
 #include <iomanip>
 #include <sstream>
+#include <string>
 #include <variant>
 
 namespace Components {
@@ -257,6 +262,97 @@ float ChatGui4::render(float time_delta) {
 
 			if (ImGui::BeginChild(chat_label.c_str(), {0, 0}, ImGuiChildFlags_Border, ImGuiWindowFlags_MenuBar)) {
 				if (ImGui::BeginMenuBar()) {
+					// check if contact has voip model
+					// use activesessioncomp instead?
+					if (_cr.all_of<VoIPModelI*>(*_selected_contact)) {
+						if (ImGui::BeginMenu("VoIP")) {
+							auto* voip_model = _cr.get<VoIPModelI*>(*_selected_contact);
+
+							std::vector<ObjectHandle> contact_sessions;
+							std::vector<ObjectHandle> acceptable_sessions;
+							for (const auto& [ov, o_vm, sc] : _os.registry().view<VoIPModelI*, Components::VoIP::SessionContact>().each()) {
+								if (o_vm != voip_model) {
+									continue;
+								}
+								if (sc.c != *_selected_contact) {
+									continue;
+								}
+
+								auto o = _os.objectHandle(ov);
+								contact_sessions.push_back(o);
+
+								if (!o.all_of<Components::VoIP::Incoming>()) {
+									continue; // not incoming
+								}
+
+								// state is ringing/not yet accepted
+								const auto* session_state = o.try_get<Components::VoIP::SessionState>();
+								if (session_state == nullptr) {
+									continue;
+								}
+
+								if (session_state->state != Components::VoIP::SessionState::State::RINGING) {
+									continue;
+								}
+								acceptable_sessions.push_back(o);
+							}
+
+							static Components::VoIP::DefaultConfig g_default_connections{};
+
+							if (ImGui::BeginMenu("default connections")) {
+								ImGui::MenuItem("incoming audio", nullptr, &g_default_connections.incoming_audio);
+								ImGui::MenuItem("incoming video", nullptr, &g_default_connections.incoming_video);
+								ImGui::Separator();
+								ImGui::MenuItem("outgoing audio", nullptr, &g_default_connections.outgoing_audio);
+								ImGui::MenuItem("outgoing video", nullptr, &g_default_connections.outgoing_video);
+								ImGui::EndMenu();
+							}
+
+							if (acceptable_sessions.size() < 2) {
+								if (ImGui::MenuItem("accept call", nullptr, false, !acceptable_sessions.empty())) {
+									voip_model->accept(acceptable_sessions.front(), g_default_connections);
+								}
+							} else {
+								if (ImGui::BeginMenu("accept call", !acceptable_sessions.empty())) {
+									for (const auto o : acceptable_sessions) {
+										std::string label = "accept #";
+										label += std::to_string(entt::to_integral(entt::to_entity(o.entity())));
+
+										if (ImGui::MenuItem(label.c_str())) {
+											voip_model->accept(o, g_default_connections);
+										}
+									}
+									ImGui::EndMenu();
+								}
+							}
+
+							// TODO: disable if already in call?
+							if (ImGui::Button(" call ")) {
+								voip_model->enter(*_selected_contact, g_default_connections);
+							}
+
+							if (contact_sessions.size() < 2) {
+								if (ImGui::MenuItem("leave/reject call", nullptr, false, !contact_sessions.empty())) {
+									voip_model->leave(contact_sessions.front());
+								}
+							} else {
+								if (ImGui::BeginMenu("leave/reject call")) {
+									// list
+									for (const auto o : contact_sessions) {
+										std::string label = "end #";
+										label += std::to_string(entt::to_integral(entt::to_entity(o.entity())));
+
+										if (ImGui::MenuItem(label.c_str())) {
+											voip_model->leave(o);
+										}
+									}
+									ImGui::EndMenu();
+								}
+							}
+
+							ImGui::EndMenu();
+						}
+					}
 					if (ImGui::BeginMenu("debug")) {
 						ImGui::Checkbox("show extra info", &_show_chat_extra_info);
 						ImGui::Checkbox("show avatar transfers", &_show_chat_avatar_tf);

src/debug_video_tap.cpp

@@ -0,0 +1,296 @@
+#include "./debug_video_tap.hpp"
+
+#include <solanaceae/object_store/object_store.hpp>
+
+#include <entt/entity/entity.hpp>
+
+#include <SDL3/SDL.h>
+
+#include <imgui/imgui.h>
+
+#include "./frame_streams/sdl/video.hpp"
+#include "./frame_streams/frame_stream2.hpp"
+
+#include <string>
+#include <memory>
+#include <mutex>
+#include <deque>
+
+#include <thread>
+#include <chrono>
+#include <atomic>
+
+#include <iostream>
+
+// fwd
+namespace Message {
+uint64_t getTimeMS(void);
+}
+
+// threadsafe queue frame stream
+// protected by a simple mutex lock
+template<typename FrameType>
+struct LockedFrameStream2 : public FrameStream2I<FrameType> {
+	std::mutex _lock;
+
+	std::deque<FrameType> _frames;
+
+	~LockedFrameStream2(void) {}
+
+	int32_t size(void) { return -1; }
+
+	std::optional<FrameType> pop(void) {
+		std::lock_guard lg{_lock};
+
+		if (_frames.empty()) {
+			return std::nullopt;
+		}
+
+		FrameType new_frame = std::move(_frames.front());
+		_frames.pop_front();
+
+		return std::move(new_frame);
+	}
+
+	bool push(const FrameType& value) {
+		std::lock_guard lg{_lock};
+
+		_frames.push_back(value);
+
+		return true;
+	}
+};
+
+struct DebugVideoTapSink : public FrameStream2SinkI<SDLVideoFrame> {
+	TextureUploaderI& _tu;
+
+	uint32_t _id_counter {0};
+
+	struct Writer {
+		struct View {
+			uint32_t _id {0}; // for stable imgui ids
+
+			uint64_t _tex {0};
+			uint32_t _tex_w {0};
+			uint32_t _tex_h {0};
+
+			bool _mirror {false}; // flip horizontally
+
+			uint64_t _v_last_ts {0}; // us
+			float _v_interval_avg {0.f}; // s
+		} view;
+
+		std::shared_ptr<LockedFrameStream2<SDLVideoFrame>> stream;
+	};
+	std::vector<Writer> _writers;
+
+	DebugVideoTapSink(TextureUploaderI& tu) : _tu(tu) {}
+	~DebugVideoTapSink(void) {}
+
+	// sink
+	std::shared_ptr<FrameStream2I<SDLVideoFrame>> subscribe(void) override {
+		_writers.emplace_back(Writer{
+			Writer::View{_id_counter++},
+			std::make_shared<LockedFrameStream2<SDLVideoFrame>>()
+		});
+
+		return _writers.back().stream;
+	}
+
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<SDLVideoFrame>>& sub) override {
+		if (!sub || _writers.empty()) {
+			// nah
+			return false;
+		}
+
+		for (auto it = _writers.cbegin(); it != _writers.cend(); it++) {
+			if (it->stream == sub) {
+				_tu.destroy(it->view._tex);
+				_writers.erase(it);
+				return true;
+			}
+		}
+
+		// what
+		return false;
+	}
+};
+
+struct DebugVideoTestSource : public FrameStream2SourceI<SDLVideoFrame> {
+	std::vector<std::shared_ptr<LockedFrameStream2<SDLVideoFrame>>> _readers;
+
+	std::atomic_bool _stop {false};
+	std::thread _thread;
+
+	DebugVideoTestSource(void) {
+		std::cout << "DVTS: starting new test video source\n";
+		_thread = std::thread([this](void) {
+			while (!_stop) {
+				if (!_readers.empty()) {
+					auto* surf = SDL_CreateSurface(960, 720, SDL_PIXELFORMAT_ARGB32);
+
+					// color
+					static auto start_time = Message::getTimeMS();
+					const float time = (Message::getTimeMS() - start_time)/1000.f;
+					SDL_ClearSurface(surf, std::sin(time), std::cos(time), 0.5f, 1.f);
+
+					SDLVideoFrame frame{ // non-owning
+						Message::getTimeMS()*1000,
+						surf,
+					};
+
+					for (auto& stream : _readers) {
+						stream->push(frame); // copy
+					}
+
+					SDL_DestroySurface(surf);
+				}
+
+				std::this_thread::sleep_for(std::chrono::milliseconds(50));
+			}
+		});
+	}
+	~DebugVideoTestSource(void) {
+		_stop = true;
+		_thread.join();
+	}
+
+	std::shared_ptr<FrameStream2I<SDLVideoFrame>> subscribe(void) override {
+		return _readers.emplace_back(std::make_shared<LockedFrameStream2<SDLVideoFrame>>());
+	}
+
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<SDLVideoFrame>>& sub) override {
+		for (auto it = _readers.cbegin(); it != _readers.cend(); it++) {
+			if (it->get() == sub.get()) {
+				_readers.erase(it);
+				return true;
+			}
+		}
+
+		return false;
+	}
+};
+
+DebugVideoTap::DebugVideoTap(ObjectStore2& os, StreamManager& sm, TextureUploaderI& tu) : _os(os), _sm(sm), _tu(tu) {
+	// post self as video sink
+	_tap = {_os.registry(), _os.registry().create()};
+	try {
+		auto dvts = std::make_unique<DebugVideoTapSink>(_tu);
+		_tap.emplace<DebugVideoTapSink*>(dvts.get()); // to get our data back
+		_tap.emplace<Components::FrameStream2Sink<SDLVideoFrame>>(
+			std::move(dvts)
+		);
+
+		_tap.emplace<Components::StreamSink>(Components::StreamSink::create<SDLVideoFrame>("DebugVideoTap"));
+		_tap.emplace<Components::TagDefaultTarget>();
+
+		_os.throwEventConstruct(_tap);
+	} catch (...) {
+		_os.registry().destroy(_tap);
+	}
+
+	_src = {_os.registry(), _os.registry().create()};
+	try {
+		auto dvts = std::make_unique<DebugVideoTestSource>();
+		_src.emplace<DebugVideoTestSource*>(dvts.get());
+		_src.emplace<Components::FrameStream2Source<SDLVideoFrame>>(
+			std::move(dvts)
+		);
+
+		_src.emplace<Components::StreamSource>(Components::StreamSource::create<SDLVideoFrame>("DebugVideoTest"));
+
+		_os.throwEventConstruct(_src);
+	} catch (...) {
+		_os.registry().destroy(_src);
+	}
+}
+
+DebugVideoTap::~DebugVideoTap(void) {
+	if (static_cast<bool>(_tap)) {
+		_os.registry().destroy(_tap);
+	}
+	if (static_cast<bool>(_src)) {
+		_os.registry().destroy(_src);
+	}
+}
+
+float DebugVideoTap::render(void) {
+	float min_interval {2.f};
+	auto& dvtsw = _tap.get<DebugVideoTapSink*>()->_writers;
+	for (auto& [view, stream] : dvtsw) {
+		std::string window_title {"DebugVideoTap #"};
+		window_title += std::to_string(view._id);
+		ImGui::SetNextWindowSize({250, 250}, ImGuiCond_Appearing);
+		if (ImGui::Begin(window_title.c_str())) {
+			while (auto new_frame_opt = stream->pop()) {
+				// timing
+				if (view._v_last_ts == 0) {
+					view._v_last_ts = new_frame_opt.value().timestampUS;
+				} else {
+					auto delta = int64_t(new_frame_opt.value().timestampUS) - int64_t(view._v_last_ts);
+					view._v_last_ts = new_frame_opt.value().timestampUS;
+
+					if (view._v_interval_avg == 0) {
+						view._v_interval_avg = delta/1'000'000.f;
+					} else {
+						const float r = 0.2f;
+						view._v_interval_avg = view._v_interval_avg * (1.f-r) + (delta/1'000'000.f) * r;
+					}
+				}
+
+				SDL_Surface* new_frame_surf = new_frame_opt.value().surface.get();
+
+				SDL_Surface* converted_surf = new_frame_surf;
+				if (new_frame_surf->format != SDL_PIXELFORMAT_RGBA32) {
+					// we need to convert
+					//std::cerr << "DVT: need to convert\n";
+					converted_surf = SDL_ConvertSurfaceAndColorspace(new_frame_surf, SDL_PIXELFORMAT_RGBA32, nullptr, SDL_COLORSPACE_RGB_DEFAULT, 0);
+					assert(converted_surf->format == SDL_PIXELFORMAT_RGBA32);
+				}
+
+				SDL_LockSurface(converted_surf);
+				if (view._tex == 0 || (int)view._tex_w != converted_surf->w || (int)view._tex_h != converted_surf->h) {
+					_tu.destroy(view._tex);
+					view._tex = _tu.uploadRGBA(
+						static_cast<const uint8_t*>(converted_surf->pixels),
+						converted_surf->w,
+						converted_surf->h,
+						TextureUploaderI::LINEAR,
+						TextureUploaderI::STREAMING
+					);
+
+					view._tex_w = converted_surf->w;
+					view._tex_h = converted_surf->h;
+				} else {
+					_tu.updateRGBA(view._tex, static_cast<const uint8_t*>(converted_surf->pixels), converted_surf->w * converted_surf->h * 4);
+				}
+				SDL_UnlockSurface(converted_surf);
+
+				if (new_frame_surf != converted_surf) {
+					// clean up temp
+					SDL_DestroySurface(converted_surf);
+				}
+			}
+
+			ImGui::Checkbox("mirror", &view._mirror);
+
+			// img here
+			if (view._tex != 0) {
+				ImGui::SameLine();
+				ImGui::Text("moving avg interval: %f", view._v_interval_avg);
+				const float img_w = ImGui::GetContentRegionAvail().x;
+				ImGui::Image(
+					reinterpret_cast<ImTextureID>(view._tex),
+					ImVec2{img_w, img_w * float(view._tex_h)/view._tex_w},
+					ImVec2{view._mirror?1.f:0.f, 0},
+					ImVec2{view._mirror?0.f:1.f, 1}
+				);
+			}
+
+		}
+		ImGui::End();
+	}
+
+	return min_interval;
+}
+

src/debug_video_tap.hpp

@@ -0,0 +1,23 @@
+#pragma once
+
+#include <solanaceae/object_store/fwd.hpp>
+#include "./frame_streams/stream_manager.hpp"
+#include "./texture_uploader.hpp"
+
+// provides a sink and a small window displaying a SDLVideoFrame
+// HACK: provides a test video source
+class DebugVideoTap {
+	ObjectStore2& _os;
+	StreamManager& _sm;
+	TextureUploaderI& _tu;
+
+	ObjectHandle _tap;
+	ObjectHandle _src;
+
+	public:
+		DebugVideoTap(ObjectStore2& os, StreamManager& sm, TextureUploaderI& tu);
+		~DebugVideoTap(void);
+
+		float render(void);
+};
+

src/frame_streams/audio_stream2.hpp

@@ -0,0 +1,39 @@
+#pragma once
+
+#include "./frame_stream2.hpp"
+
+#include <solanaceae/util/span.hpp>
+
+#include <cstdint>
+#include <variant>
+#include <vector>
+
+// raw audio
+// channels make samples interleaved,
+// planar channels are not supported
+// s16 only stopgap audio frame (simplified)
+struct AudioFrame2 {
+	// samples per second
+	uint32_t sample_rate {48'000};
+
+	// only >0 is valid
+	size_t channels {0};
+
+	std::variant<
+		std::vector<int16_t>, // S16, platform endianess
+		Span<int16_t> // non owning variant, for direct consumption
+	> buffer;
+
+	// helpers
+	Span<int16_t> getSpan(void) const {
+		if (std::holds_alternative<std::vector<int16_t>>(buffer)) {
+			return Span<int16_t>{std::get<std::vector<int16_t>>(buffer)};
+		} else {
+			return std::get<Span<int16_t>>(buffer);
+		}
+		return {};
+	}
+};
+
+using AudioFrame2Stream2I = FrameStream2I<AudioFrame2>;
+

src/frame_streams/audio_stream_pop_reframer.hpp

@@ -0,0 +1,103 @@
+#pragma once
+
+#include "./audio_stream2.hpp"
+
+// reframes audio frames to a specified size in ms
+// TODO: use absolute sample count instead??
+template<typename RealAudioStream>
+struct AudioStreamPopReFramer : public FrameStream2I<AudioFrame2> {
+	uint32_t _frame_length_ms {20};
+
+	// gotta be careful of the multithreaded nature
+	// and false(true) sharing
+	uint64_t _pad0{};
+	RealAudioStream _stream;
+	uint64_t _pad1{};
+
+	// dequeue?
+	std::vector<int16_t> _buffer;
+
+	uint32_t _sample_rate {48'000};
+	size_t _channels {0};
+
+	AudioStreamPopReFramer(uint32_t frame_length_ms = 20)
+	: _frame_length_ms(frame_length_ms) {
+	}
+
+	AudioStreamPopReFramer(uint32_t frame_length_ms, FrameStream2I<AudioFrame2>&& stream)
+	: _frame_length_ms(frame_length_ms), _stream(std::move(stream)) {
+	}
+
+	~AudioStreamPopReFramer(void) {}
+
+	size_t getDesiredSize(void) const {
+		return _frame_length_ms * _sample_rate * _channels / 1000;
+	}
+
+	int32_t size(void) override { return -1; }
+
+	std::optional<AudioFrame2> pop(void) override {
+		do {
+			auto new_in = _stream.pop();
+			if (new_in.has_value()) {
+				auto& new_value = new_in.value();
+
+				// changed
+				if (_sample_rate != new_value.sample_rate || _channels != new_value.channels) {
+					//if (_channels != 0) {
+					//    std::cerr << "ReFramer warning: reconfiguring, dropping buffer\n";
+					//}
+					_sample_rate = new_value.sample_rate;
+					_channels = new_value.channels;
+
+					// buffer does not exist or config changed and we discard
+					_buffer = {};
+				}
+
+				//std::cout << "new incoming frame is " << new_value.getSpan().size/new_value.channels*1000/new_value.sample_rate << "ms\n";
+
+				auto new_span = new_value.getSpan();
+
+				if (_buffer.empty()) {
+					_buffer = {new_span.cbegin(), new_span.cend()};
+				} else {
+					_buffer.insert(_buffer.cend(), new_span.cbegin(), new_span.cend());
+				}
+			} else if (_buffer.empty()) {
+				// first pop might result in invalid state
+				return std::nullopt;
+			} else {
+				// inner stream pop did not give a new value
+				break; // out of loop
+			}
+		} while (_buffer.size() < getDesiredSize());
+
+		const auto desired_size = getDesiredSize();
+
+		// > threshold?
+		if (_buffer.size() < desired_size) {
+			return std::nullopt;
+		}
+
+		// copy data
+		std::vector<int16_t> return_buffer(_buffer.cbegin(), _buffer.cbegin()+desired_size);
+
+		// now crop buffer (meh)
+		// move data from back to front
+		_buffer.erase(_buffer.cbegin(), _buffer.cbegin() + desired_size);
+
+		return AudioFrame2{
+			_sample_rate,
+			_channels,
+			std::move(return_buffer),
+		};
+	}
+
+	bool push(const AudioFrame2& value) override {
+		// might be worth it to instead do the work on push
+		//assert(false && "push reframing not implemented");
+		// passthrough
+		return _stream.push(value);
+	}
+};
+

src/frame_streams/frame_stream2.hpp

@@ -0,0 +1,47 @@
+#pragma once
+
+#include <cstdint>
+#include <memory>
+#include <optional>
+#include <vector>
+
+// Frames often consist of:
+// - seq id // incremental sequential id, gaps in ids can be used to detect loss
+// - or timestamp
+// - data // the frame data
+// eg:
+//struct ExampleFrame {
+	//int64_t seq_id {0};
+	//std::vector<uint8_t> data;
+//};
+
+template<typename FrameType>
+struct FrameStream2I {
+	virtual ~FrameStream2I(void) {}
+
+	// get number of available frames
+	// returns -1 if unknown
+	[[nodiscard]] virtual int32_t size(void) = 0;
+
+	// get next frame
+	// data sharing? -> no, data is copied for each fsr, if concurency supported
+	[[nodiscard]] virtual std::optional<FrameType> pop(void) = 0;
+
+	// returns true if there are readers (or we dont know)
+	virtual bool push(const FrameType& value) = 0;
+};
+
+template<typename FrameType>
+struct FrameStream2SourceI {
+	virtual ~FrameStream2SourceI(void) {}
+	[[nodiscard]] virtual std::shared_ptr<FrameStream2I<FrameType>> subscribe(void) = 0;
+	virtual bool unsubscribe(const std::shared_ptr<FrameStream2I<FrameType>>& sub) = 0;
+};
+
+template<typename FrameType>
+struct FrameStream2SinkI {
+	virtual ~FrameStream2SinkI(void) {}
+	[[nodiscard]] virtual std::shared_ptr<FrameStream2I<FrameType>> subscribe(void) = 0;
+	virtual bool unsubscribe(const std::shared_ptr<FrameStream2I<FrameType>>& sub) = 0;
+};
+

src/frame_streams/locked_frame_stream.hpp

@@ -0,0 +1,46 @@
+#pragma once
+
+#include "./frame_stream2.hpp"
+
+#include <mutex>
+#include <deque>
+
+// threadsafe queue frame stream
+// protected by a simple mutex lock
+// prefer lockless queue implementations, when available
+template<typename FrameType>
+struct LockedFrameStream2 : public FrameStream2I<FrameType> {
+	std::mutex _lock;
+
+	std::deque<FrameType> _frames;
+
+	~LockedFrameStream2(void) {}
+
+	int32_t size(void) { return -1; }
+
+	std::optional<FrameType> pop(void) {
+		std::lock_guard lg{_lock};
+
+		if (_frames.empty()) {
+			return std::nullopt;
+		}
+
+		FrameType new_frame = std::move(_frames.front());
+		_frames.pop_front();
+
+		return std::move(new_frame);
+	}
+
+	bool push(const FrameType& value) {
+		std::lock_guard lg{_lock};
+
+		if (_frames.size() > 1024) {
+			return false; // hard limit
+		}
+
+		_frames.push_back(value);
+
+		return true;
+	}
+};
+

src/frame_streams/multi_source.hpp

@@ -0,0 +1,62 @@
+#pragma once
+
+#include "./locked_frame_stream.hpp"
+
+#include <cassert>
+
+// implements a stream that pushes to all sub streams
+template<typename FrameType, typename SubStreamType = LockedFrameStream2<FrameType>>
+struct FrameStream2MultiSource : public FrameStream2SourceI<FrameType>, public FrameStream2I<FrameType> {
+	using sub_stream_type_t = SubStreamType;
+
+	// pointer stability
+	std::vector<std::shared_ptr<SubStreamType>> _sub_streams;
+	std::mutex _sub_stream_lock; // accessing the _sub_streams array needs to be exclusive
+	// a simple lock here is ok, since this tends to be a rare operation,
+	// except for the push, which is always on the same thread
+
+	virtual ~FrameStream2MultiSource(void) {}
+
+	// TODO: forward args instead
+	std::shared_ptr<FrameStream2I<FrameType>> subscribe(void) override {
+		std::lock_guard lg{_sub_stream_lock};
+		return _sub_streams.emplace_back(std::make_unique<SubStreamType>());
+	}
+
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<FrameType>>& sub) override {
+		std::lock_guard lg{_sub_stream_lock};
+		for (auto it = _sub_streams.begin(); it != _sub_streams.end(); it++) {
+			if (*it == sub) {
+				_sub_streams.erase(it);
+				return true;
+			}
+		}
+		return false; // ?
+	}
+
+	// stream interface
+
+	int32_t size(void) override {
+		// TODO: return something sensible?
+		return -1;
+	}
+
+	std::optional<FrameType> pop(void) override {
+		// nope
+		assert(false && "this logic is very frame type specific, provide an impl");
+		return std::nullopt;
+	}
+
+	// returns true if there are readers
+	bool push(const FrameType& value) override {
+		std::lock_guard lg{_sub_stream_lock};
+		bool have_readers{false};
+		for (auto& it : _sub_streams) {
+			[[maybe_unused]] auto _ = it->push(value);
+			have_readers = true; // even if queue full, we still continue believing in them
+			// maybe consider push return value?
+		}
+		return have_readers;
+	}
+};
+

src/frame_streams/sdl/sdl_audio2_frame_stream2.cpp

@@ -0,0 +1,279 @@
+#include "./sdl_audio2_frame_stream2.hpp"
+
+#include <cassert>
+#include <iostream>
+#include <optional>
+
+#include "../audio_stream_pop_reframer.hpp"
+
+// "thin" wrapper around sdl audio streams
+// we dont needs to get fance, as they already provide everything we need
+struct SDLAudio2StreamReader : public AudioFrame2Stream2I {
+	std::unique_ptr<SDL_AudioStream, decltype(&SDL_DestroyAudioStream)> _stream;
+
+	uint32_t _sample_rate {48'000};
+	size_t _channels {0};
+
+	// buffer gets reused!
+	std::vector<int16_t> _buffer;
+
+	SDLAudio2StreamReader(void) : _stream(nullptr, nullptr) {}
+	SDLAudio2StreamReader(SDLAudio2StreamReader&& other) :
+		_stream(std::move(other._stream)),
+		_sample_rate(other._sample_rate),
+		_channels(other._channels)
+	{
+		const size_t buffer_size {960*_channels};
+		_buffer.resize(buffer_size);
+	}
+
+	~SDLAudio2StreamReader(void) {
+		if (_stream) {
+			SDL_UnbindAudioStream(_stream.get());
+		}
+	}
+
+	int32_t size(void) override {
+		//assert(_stream);
+		// returns bytes
+		//SDL_GetAudioStreamAvailable(_stream.get());
+		return -1;
+	}
+
+	std::optional<AudioFrame2> pop(void) override {
+		assert(_stream);
+		if (!_stream) {
+			return std::nullopt;
+		}
+
+		const size_t buffer_size {960*_channels};
+		_buffer.resize(buffer_size); // noop?
+
+		const auto read_bytes = SDL_GetAudioStreamData(
+			_stream.get(),
+			_buffer.data(),
+			_buffer.size()*sizeof(int16_t)
+		);
+
+		// no new frame yet, or error
+		if (read_bytes <= 0) {
+			return std::nullopt;
+		}
+
+		return AudioFrame2{
+			_sample_rate, _channels,
+			Span<int16_t>(_buffer.data(), read_bytes/sizeof(int16_t)),
+		};
+	}
+
+	bool push(const AudioFrame2&) override {
+		// TODO: make universal sdl stream wrapper (combine with SDLAudioOutputDeviceDefaultInstance)
+		assert(false && "read only");
+		return false;
+	}
+};
+
+SDLAudio2InputDevice::SDLAudio2InputDevice(void) : SDLAudio2InputDevice(SDL_AUDIO_DEVICE_DEFAULT_RECORDING) {
+}
+
+SDLAudio2InputDevice::SDLAudio2InputDevice(SDL_AudioDeviceID conf_device_id) : _configured_device_id(conf_device_id) {
+	if (_configured_device_id == 0) {
+		// TODO: proper error handling
+		throw int(1);
+	}
+}
+
+SDLAudio2InputDevice::~SDLAudio2InputDevice(void) {
+	_streams.clear();
+
+	if (_virtual_device_id != 0) {
+		SDL_CloseAudioDevice(_virtual_device_id);
+		_virtual_device_id = 0;
+	}
+}
+
+std::shared_ptr<FrameStream2I<AudioFrame2>> SDLAudio2InputDevice::subscribe(void) {
+	if (_virtual_device_id == 0) {
+		// first stream, open device
+		// this spec is more like a hint to the hardware
+		SDL_AudioSpec spec {
+			SDL_AUDIO_S16,
+			1, // TODO: conf
+			48'000,
+		};
+		_virtual_device_id = SDL_OpenAudioDevice(_configured_device_id, &spec);
+	}
+
+	if (_virtual_device_id == 0) {
+		std::cerr << "SDLAID error: failed opening device " << _configured_device_id << "\n";
+		return nullptr;
+	}
+
+	SDL_AudioSpec spec {
+		SDL_AUDIO_S16, // required, as AudioFrame2 only supports s16
+		1, // TODO: conf
+		48'000,
+	};
+
+	SDL_AudioSpec device_spec {
+		SDL_AUDIO_S16,
+		1, // TODO: conf
+		48'000,
+	};
+	// TODO: error check
+	SDL_GetAudioDeviceFormat(_virtual_device_id, &device_spec, nullptr);
+
+	// error check
+	auto* sdl_stream = SDL_CreateAudioStream(&device_spec, &spec);
+
+	// error check
+	SDL_BindAudioStream(_virtual_device_id, sdl_stream);
+
+	//auto new_stream = std::make_shared<SDLAudio2StreamReader>();
+	//// TODO: move to ctr
+	//new_stream->_stream = {sdl_stream, &SDL_DestroyAudioStream};
+	//new_stream->_sample_rate = spec.freq;
+	//new_stream->_channels = spec.channels;
+
+	auto new_stream = std::make_shared<AudioStreamPopReFramer<SDLAudio2StreamReader>>();
+	new_stream->_stream._stream = {sdl_stream, &SDL_DestroyAudioStream};
+	new_stream->_stream._sample_rate = spec.freq;
+	new_stream->_stream._channels = spec.channels;
+	new_stream->_frame_length_ms = 5; // WHY DOES THIS FIX MY ISSUE !!!
+
+	_streams.emplace_back(new_stream);
+
+	return new_stream;
+}
+
+bool SDLAudio2InputDevice::unsubscribe(const std::shared_ptr<FrameStream2I<AudioFrame2>>& sub) {
+	for (auto it = _streams.cbegin(); it != _streams.cend(); it++) {
+		if (*it == sub) {
+			_streams.erase(it);
+			if (_streams.empty()) {
+				// last stream, close
+				// TODO: make sure no shared ptr still exists???
+				SDL_CloseAudioDevice(_virtual_device_id);
+				std::cout << "SDLAID: closing device " << _virtual_device_id << " (" << _configured_device_id << ")\n";
+				_virtual_device_id = 0;
+			}
+			return true;
+		}
+	}
+
+	return false;
+}
+
+// does not need to be visible in the header
+struct SDLAudio2OutputDeviceDefaultInstance : public AudioFrame2Stream2I {
+	std::unique_ptr<SDL_AudioStream, decltype(&SDL_DestroyAudioStream)> _stream;
+
+	uint32_t _last_sample_rate {48'000};
+	size_t _last_channels {0};
+
+	// TODO: audio device
+	SDLAudio2OutputDeviceDefaultInstance(void);
+	SDLAudio2OutputDeviceDefaultInstance(SDLAudio2OutputDeviceDefaultInstance&& other);
+
+	~SDLAudio2OutputDeviceDefaultInstance(void);
+
+	int32_t size(void) override;
+	std::optional<AudioFrame2> pop(void) override;
+	bool push(const AudioFrame2& value) override;
+};
+
+SDLAudio2OutputDeviceDefaultInstance::SDLAudio2OutputDeviceDefaultInstance(void) : _stream(nullptr, nullptr) {
+}
+
+SDLAudio2OutputDeviceDefaultInstance::SDLAudio2OutputDeviceDefaultInstance(SDLAudio2OutputDeviceDefaultInstance&& other) : _stream(std::move(other._stream)) {
+}
+
+SDLAudio2OutputDeviceDefaultInstance::~SDLAudio2OutputDeviceDefaultInstance(void) {
+}
+int32_t SDLAudio2OutputDeviceDefaultInstance::size(void) {
+	return -1;
+}
+
+std::optional<AudioFrame2> SDLAudio2OutputDeviceDefaultInstance::pop(void) {
+	assert(false);
+	// this is an output device, there is no data to pop
+	return std::nullopt;
+}
+
+bool SDLAudio2OutputDeviceDefaultInstance::push(const AudioFrame2& value) {
+	if (!static_cast<bool>(_stream)) {
+		return false;
+	}
+
+	// verify here the fame has the same channel count and sample freq
+	// if something changed, we need to either use a temporary stream, just for conversion, or reopen _stream with the new params
+	// because of data temporality, the second options looks like a better candidate
+	if (
+		value.sample_rate != _last_sample_rate ||
+		value.channels != _last_channels
+	) {
+		const SDL_AudioSpec spec = {
+			static_cast<SDL_AudioFormat>(SDL_AUDIO_S16),
+			static_cast<int>(value.channels),
+			static_cast<int>(value.sample_rate)
+		};
+
+		SDL_SetAudioStreamFormat(_stream.get(), &spec, nullptr);
+
+		std::cerr << "SDLAOD: audio format changed\n";
+	}
+
+	auto data = value.getSpan();
+
+	if (data.size == 0) {
+		std::cerr << "empty audio frame??\n";
+	}
+
+	if (!SDL_PutAudioStreamData(_stream.get(), data.ptr, data.size * sizeof(int16_t))) {
+		std::cerr << "put data error\n";
+		return false; // return true?
+	}
+
+	_last_sample_rate = value.sample_rate;
+	_last_channels = value.channels;
+
+	return true;
+}
+
+
+SDLAudio2OutputDeviceDefaultSink::~SDLAudio2OutputDeviceDefaultSink(void) {
+	// TODO: pause and close device?
+}
+
+std::shared_ptr<FrameStream2I<AudioFrame2>> SDLAudio2OutputDeviceDefaultSink::subscribe(void) {
+	auto new_instance = std::make_shared<SDLAudio2OutputDeviceDefaultInstance>();
+
+	constexpr SDL_AudioSpec spec = { SDL_AUDIO_S16, 2, 48000 };
+
+	new_instance->_stream = {
+		SDL_OpenAudioDeviceStream(SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK, &spec, nullptr, nullptr),
+		&SDL_DestroyAudioStream
+	};
+	new_instance->_last_sample_rate = spec.freq;
+	new_instance->_last_channels = spec.channels;
+
+	if (!static_cast<bool>(new_instance->_stream)) {
+		std::cerr << "SDL open audio device failed!\n";
+		return nullptr;
+	}
+
+	const auto audio_device_id = SDL_GetAudioStreamDevice(new_instance->_stream.get());
+	SDL_ResumeAudioDevice(audio_device_id);
+
+	return new_instance;
+}
+
+bool SDLAudio2OutputDeviceDefaultSink::unsubscribe(const std::shared_ptr<FrameStream2I<AudioFrame2>>& sub) {
+	// TODO: i think we should keep track of them
+	if (!sub) {
+		return false;
+	}
+
+	return true;
+}
+

src/frame_streams/sdl/sdl_audio2_frame_stream2.hpp

@@ -0,0 +1,43 @@
+#pragma once
+
+#include "../frame_stream2.hpp"
+#include "../audio_stream2.hpp"
+
+#include <SDL3/SDL.h>
+
+#include <cstdint>
+#include <vector>
+
+// we dont have to multicast ourself, because sdl streams and virtual devices already do this
+
+// source
+// opens device
+// creates a sdl audio stream for each subscribed reader stream
+struct SDLAudio2InputDevice : public FrameStream2SourceI<AudioFrame2> {
+	// held by instances
+	using sdl_stream_type = std::unique_ptr<SDL_AudioStream, decltype(&SDL_DestroyAudioStream)>;
+
+	SDL_AudioDeviceID _configured_device_id {0};
+	SDL_AudioDeviceID _virtual_device_id {0};
+
+	std::vector<std::shared_ptr<FrameStream2I<AudioFrame2>>> _streams;
+
+	SDLAudio2InputDevice(void);
+	SDLAudio2InputDevice(SDL_AudioDeviceID conf_device_id);
+	~SDLAudio2InputDevice(void);
+
+	std::shared_ptr<FrameStream2I<AudioFrame2>> subscribe(void) override;
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<AudioFrame2>>& sub) override;
+};
+
+// sink
+// constructs entirely new streams, since sdl handles sync and mixing for us (or should)
+struct SDLAudio2OutputDeviceDefaultSink : public FrameStream2SinkI<AudioFrame2> {
+	// TODO: pause device?
+
+	~SDLAudio2OutputDeviceDefaultSink(void);
+
+	std::shared_ptr<FrameStream2I<AudioFrame2>> subscribe(void) override;
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<AudioFrame2>>& sub) override;
+};
+

src/frame_streams/sdl/video.hpp

@@ -0,0 +1,41 @@
+#pragma once
+
+#include <SDL3/SDL.h>
+
+#include <cstdint>
+#include <memory>
+
+// https://youtu.be/71Iw4Q74OaE
+
+inline void nopSurfaceDestructor(SDL_Surface*) {}
+
+// this is very sdl specific
+// but allows us to autoconvert between formats (to a degree)
+struct SDLVideoFrame {
+	// micro seconds (nano is way too much)
+	uint64_t timestampUS {0};
+
+	std::unique_ptr<SDL_Surface, decltype(&SDL_DestroySurface)> surface {nullptr, &SDL_DestroySurface};
+
+	// special non-owning constructor
+	SDLVideoFrame(
+		uint64_t ts,
+		SDL_Surface* surf
+	) {
+		timestampUS = ts;
+		surface = {surf, &nopSurfaceDestructor};
+	}
+	SDLVideoFrame(SDLVideoFrame&& other) = default;
+	// copy
+	SDLVideoFrame(const SDLVideoFrame& other) {
+		timestampUS = other.timestampUS;
+		if (static_cast<bool>(other.surface)) {
+			surface = {
+				SDL_DuplicateSurface(other.surface.get()),
+				&SDL_DestroySurface
+			};
+		}
+	}
+	SDLVideoFrame& operator=(const SDLVideoFrame& other) = delete;
+};
+

src/frame_streams/sdl/video_push_converter.cpp

@@ -0,0 +1,58 @@
+#include "./video_push_converter.hpp"
+
+SDL_Surface* convertYUY2_IYUV(SDL_Surface* surf) {
+	if (surf->format != SDL_PIXELFORMAT_YUY2) {
+		return nullptr;
+	}
+	if ((surf->w % 2) != 0) {
+		SDL_SetError("YUY2->IYUV does not support odd widths");
+		// hmmm, we dont handle odd widths
+		return nullptr;
+	}
+
+	SDL_LockSurface(surf);
+
+	SDL_Surface* conv_surf = SDL_CreateSurface(surf->w, surf->h, SDL_PIXELFORMAT_IYUV);
+	SDL_LockSurface(conv_surf);
+
+	// YUY2 is 4:2:2 packed
+	// Y is simple, we just copy it over
+	// U V are double the resolution (vertically), so we avg both
+	// Packed mode: Y0+U0+Y1+V0 (1 plane)
+
+	uint8_t* y_plane = static_cast<uint8_t*>(conv_surf->pixels);
+	uint8_t* u_plane = static_cast<uint8_t*>(conv_surf->pixels) + conv_surf->w*conv_surf->h;
+	uint8_t* v_plane = static_cast<uint8_t*>(conv_surf->pixels) + conv_surf->w*conv_surf->h + (conv_surf->w/2)*(conv_surf->h/2);
+
+	const uint8_t* yuy2_data = static_cast<const uint8_t*>(surf->pixels);
+
+	for (int y = 0; y < surf->h; y++) {
+		for (int x = 0; x < surf->w; x += 2) {
+			// every pixel uses 2 bytes
+			const uint8_t* yuy2_curser = yuy2_data + y*surf->w*2 + x*2;
+			uint8_t src_y0 = yuy2_curser[0];
+			uint8_t src_u = yuy2_curser[1];
+			uint8_t src_y1 = yuy2_curser[2];
+			uint8_t src_v = yuy2_curser[3];
+
+			y_plane[y*conv_surf->w + x] = src_y0;
+			y_plane[y*conv_surf->w + x+1] = src_y1;
+
+			size_t uv_index = (y/2) * (conv_surf->w/2) + x/2;
+			if (y % 2 == 0) {
+				// first write
+				u_plane[uv_index] = src_u;
+				v_plane[uv_index] = src_v;
+			} else {
+				// second write, mix with existing value
+				u_plane[uv_index] = (int(u_plane[uv_index]) + int(src_u)) / 2;
+				v_plane[uv_index] = (int(v_plane[uv_index]) + int(src_v)) / 2;
+			}
+		}
+	}
+
+	SDL_UnlockSurface(conv_surf);
+	SDL_UnlockSurface(surf);
+	return conv_surf;
+}
+

src/frame_streams/sdl/video_push_converter.hpp

@@ -0,0 +1,89 @@
+#pragma once
+
+#include "./video.hpp"
+#include "../frame_stream2.hpp"
+
+#include <cassert>
+
+#include <iostream> // meh
+
+static bool isFormatYUV(SDL_PixelFormat f) {
+	return
+		f == SDL_PIXELFORMAT_YV12 ||
+		f == SDL_PIXELFORMAT_IYUV ||
+		f == SDL_PIXELFORMAT_YUY2 ||
+		f == SDL_PIXELFORMAT_UYVY ||
+		f == SDL_PIXELFORMAT_YVYU ||
+		f == SDL_PIXELFORMAT_NV12 ||
+		f == SDL_PIXELFORMAT_NV21 ||
+		f == SDL_PIXELFORMAT_P010
+	;
+}
+
+SDL_Surface* convertYUY2_IYUV(SDL_Surface* surf);
+
+template<typename RealStream>
+struct PushConversionVideoStream : public RealStream {
+	SDL_PixelFormat _forced_format {SDL_PIXELFORMAT_IYUV};
+
+	template<typename... Args>
+	PushConversionVideoStream(SDL_PixelFormat forced_format, Args&&... args) : RealStream(std::forward<Args>(args)...), _forced_format(forced_format) {}
+	~PushConversionVideoStream(void) {}
+
+	bool push(const SDLVideoFrame& value) override {
+		SDL_Surface* surf = value.surface.get();
+		if (surf->format != _forced_format) {
+			//std::cerr << "DTC: need to convert from " << SDL_GetPixelFormatName(converted_surf->format) << " to SDL_PIXELFORMAT_IYUV\n";
+			if (surf->format == SDL_PIXELFORMAT_YUY2 && _forced_format == SDL_PIXELFORMAT_IYUV) {
+				// optimized custom impl
+
+				//auto start = Message::getTimeMS();
+				surf = convertYUY2_IYUV(surf);
+				//auto end = Message::getTimeMS();
+				// 3ms
+				//std::cerr << "DTC: timing " << SDL_GetPixelFormatName(converted_surf->format) << "->SDL_PIXELFORMAT_IYUV: " << end-start << "ms\n";
+			} else if (isFormatYUV(surf->format)) {
+				// TODO: fix sdl rgb->yuv conversion resulting in too dark (colorspace) issues
+				// https://github.com/libsdl-org/SDL/issues/10877
+
+				// meh, need to convert to rgb as a stopgap
+
+				//auto start = Message::getTimeMS();
+				SDL_Surface* tmp_conv_surf = SDL_ConvertSurfaceAndColorspace(surf, SDL_PIXELFORMAT_RGB24, nullptr, SDL_COLORSPACE_RGB_DEFAULT, 0);
+				//auto end = Message::getTimeMS();
+				// 1ms
+				//std::cerr << "DTC: timing " << SDL_GetPixelFormatName(converted_surf->format) << "->SDL_PIXELFORMAT_RGB24: " << end-start << "ms\n";
+
+				//start = Message::getTimeMS();
+				surf = SDL_ConvertSurfaceAndColorspace(tmp_conv_surf, _forced_format, nullptr, SDL_COLORSPACE_YUV_DEFAULT, 0);
+				//end = Message::getTimeMS();
+				// 60ms
+				//std::cerr << "DTC: timing SDL_PIXELFORMAT_RGB24->" << SDL_GetPixelFormatName(_forced_format) << ": " << end-start << "ms\n";
+
+				SDL_DestroySurface(tmp_conv_surf);
+			} else {
+				surf = SDL_ConvertSurface(surf, _forced_format);
+			}
+
+			if (surf == nullptr) {
+				// oh god
+				std::cerr << "DTC error: failed to convert surface to IYUV: " << SDL_GetError() << "\n";
+				return false;
+			}
+		}
+		assert(surf != nullptr);
+		if (surf != value.surface.get()) {
+			// TODO: add ctr with uptr
+			SDLVideoFrame new_value{value.timestampUS, nullptr};
+			new_value.surface = {
+				surf,
+				&SDL_DestroySurface
+			};
+
+			return RealStream::push(std::move(new_value));
+		} else {
+			return RealStream::push(value);
+		}
+	}
+};
+

src/frame_streams/stream_manager.cpp

@@ -0,0 +1,206 @@
+#include "./stream_manager.hpp"
+
+StreamManager::Connection::Connection(
+	ObjectHandle src_,
+	ObjectHandle sink_,
+	std::unique_ptr<Data>&& data_,
+	std::function<void(Connection&)>&& pump_fn_,
+	std::function<void(Connection&)>&& unsubscribe_fn_,
+	bool on_main_thread_
+) :
+	src(src_),
+	sink(sink_),
+	data(std::move(data_)),
+	pump_fn(std::move(pump_fn_)),
+	unsubscribe_fn(std::move(unsubscribe_fn_)),
+	on_main_thread(on_main_thread_)
+{
+	if (!on_main_thread) {
+		// start thread
+		pump_thread = std::thread([this](void) {
+			while (!stop) {
+				pump_fn(*this);
+				std::this_thread::sleep_for(std::chrono::milliseconds(5));
+			}
+			finished = true;
+		});
+	}
+}
+
+StreamManager::StreamManager(ObjectStore2& os) : _os(os) {
+	_os.subscribe(this, ObjectStore_Event::object_construct);
+	//_os.subscribe(this, ObjectStore_Event::object_update);
+	_os.subscribe(this, ObjectStore_Event::object_destroy);
+}
+
+StreamManager::~StreamManager(void) {
+	// stop all connetions
+	for (const auto& con : _connections) {
+		con->stop = true;
+		if (!con->on_main_thread) {
+			con->pump_thread.join(); // we skip the finished check and wait
+		}
+		con->unsubscribe_fn(*con);
+	}
+}
+
+bool StreamManager::connect(Object src, Object sink, bool threaded) {
+	auto h_src = _os.objectHandle(src);
+	auto h_sink = _os.objectHandle(sink);
+	if (!static_cast<bool>(h_src) || !static_cast<bool>(h_sink)) {
+		// an object does not exist
+		return false;
+	}
+
+	// get src and sink comps
+	if (!h_src.all_of<Components::StreamSource>()) {
+		// src not stream source
+		return false;
+	}
+
+	if (!h_sink.all_of<Components::StreamSink>()) {
+		// sink not stream sink
+		return false;
+	}
+
+	const auto& ssrc = h_src.get<Components::StreamSource>();
+	const auto& ssink = h_sink.get<Components::StreamSink>();
+
+	// compare type
+	if (ssrc.frame_type_name != ssink.frame_type_name) {
+		return false;
+	}
+
+	// always fail in debug mode
+	assert(static_cast<bool>(ssrc.connect_fn));
+	if (!static_cast<bool>(ssrc.connect_fn)) {
+		return false;
+	}
+
+	// use connect fn from src
+	return ssrc.connect_fn(*this, src, sink, threaded);
+}
+
+bool StreamManager::disconnect(Object src, Object sink) {
+	auto res = std::find_if(
+		_connections.cbegin(), _connections.cend(),
+		[&](const auto& a) { return a->src == src && a->sink == sink; }
+	);
+	if (res == _connections.cend()) {
+		// not found
+		return false;
+	}
+
+	// do disconnect
+	(*res)->stop = true;
+
+	return true;
+}
+
+bool StreamManager::disconnectAll(Object o) {
+	bool succ {false};
+	for (const auto& con : _connections) {
+		if (con->src == o || con->sink == o) {
+			con->stop = true;
+			succ = true;
+		}
+	}
+
+	return succ;
+}
+
+// do we need the time delta?
+float StreamManager::tick(float) {
+	// pump all mainthread connections
+	for (auto it = _connections.begin(); it != _connections.end();) {
+		auto& con = **it;
+
+		if (!static_cast<bool>(con.src) || !static_cast<bool>(con.sink)) {
+			// either side disappeard without disconnectAll
+			// TODO: warn/error log
+			con.stop = true;
+		}
+
+		if (con.on_main_thread) {
+			con.pump_fn(con);
+		}
+
+		if (con.stop && (con.finished || con.on_main_thread)) {
+			if (!con.on_main_thread) {
+				assert(con.pump_thread.joinable());
+				con.pump_thread.join();
+			}
+			con.unsubscribe_fn(con);
+			it = _connections.erase(it);
+		} else {
+			it++;
+		}
+	}
+
+	// return min over intervals instead
+	return 2.f; // TODO: 2sec makes mainthread connections unusable
+}
+
+bool StreamManager::onEvent(const ObjectStore::Events::ObjectConstruct& e) {
+	if (!e.e.any_of<Components::StreamSink, Components::StreamSource>()) {
+		return false;
+	}
+
+	// update default targets
+	if (e.e.all_of<Components::TagDefaultTarget>()) {
+		if (e.e.all_of<Components::StreamSource>()) {
+			_default_sources[e.e.get<Components::StreamSource>().frame_type_name] = e.e;
+		} else { // sink
+			_default_sinks[e.e.get<Components::StreamSink>().frame_type_name] = e.e;
+		}
+	}
+
+	// connect to default
+	// only ever do this on new objects
+	if (e.e.all_of<Components::TagConnectToDefault>()) {
+		if (e.e.all_of<Components::StreamSource>()) {
+			auto it_d_sink = _default_sinks.find(e.e.get<Components::StreamSource>().frame_type_name);
+			if (it_d_sink != _default_sinks.cend()) {
+				// TODO: threaded
+				connect(e.e, it_d_sink->second);
+			}
+		} else { // sink
+			auto it_d_src = _default_sources.find(e.e.get<Components::StreamSink>().frame_type_name);
+			if (it_d_src != _default_sources.cend()) {
+				// TODO: threaded
+				connect(it_d_src->second, e.e);
+			}
+		}
+	}
+
+	return false;
+}
+
+bool StreamManager::onEvent(const ObjectStore::Events::ObjectUpdate&) {
+	// what do we do here?
+	return false;
+}
+
+bool StreamManager::onEvent(const ObjectStore::Events::ObjectDestory& e) {
+	// typeless
+	for (auto it = _default_sources.cbegin(); it != _default_sources.cend();) {
+		if (it->second == e.e) {
+			it = _default_sources.erase(it);
+		} else {
+			it++;
+		}
+	}
+	for (auto it = _default_sinks.cbegin(); it != _default_sinks.cend();) {
+		if (it->second == e.e) {
+			it = _default_sinks.erase(it);
+		} else {
+			it++;
+		}
+	}
+
+	// TODO: destroy connections
+	// TODO: auto reconnect default following devices if another default exists
+
+	return false;
+}
+

src/frame_streams/stream_manager.hpp

@@ -0,0 +1,222 @@
+#pragma once
+
+#include <solanaceae/object_store/fwd.hpp>
+#include <solanaceae/object_store/object_store.hpp>
+
+#include <entt/core/type_info.hpp>
+
+#include "./frame_stream2.hpp"
+
+#include <unordered_map>
+#include <vector>
+#include <memory>
+#include <algorithm>
+#include <thread>
+#include <chrono>
+#include <atomic>
+
+// fwd
+class StreamManager;
+
+namespace Components {
+
+	// mark a source or sink as the(a) default
+	struct TagDefaultTarget {};
+
+	// mark a source/sink as to be connected to a default sink/source
+	// of the same type
+	struct TagConnectToDefault {};
+
+	struct StreamSource {
+		std::string name;
+		std::string frame_type_name;
+
+		std::function<bool(StreamManager&, Object, Object, bool)> connect_fn;
+
+		template<typename FrameType>
+		static StreamSource create(const std::string& name);
+	};
+
+	struct StreamSink {
+		std::string name;
+		std::string frame_type_name;
+
+		template<typename FrameType>
+		static StreamSink create(const std::string& name);
+	};
+
+	template<typename FrameType>
+	using FrameStream2Source = std::unique_ptr<FrameStream2SourceI<FrameType>>;
+
+	template<typename FrameType>
+	using FrameStream2Sink = std::unique_ptr<FrameStream2SinkI<FrameType>>;
+
+} // Components
+
+
+class StreamManager : protected ObjectStoreEventI {
+	friend class StreamManagerUI; // TODO: make this go away
+	ObjectStore2& _os;
+
+	struct Connection {
+		ObjectHandle src;
+		ObjectHandle sink;
+
+		struct Data {
+			virtual ~Data(void) {}
+		};
+		std::unique_ptr<Data> data; // stores reader writer type erased
+		std::function<void(Connection&)> pump_fn; // TODO: make it return next interval?
+		std::function<void(Connection&)> unsubscribe_fn;
+
+		bool on_main_thread {true};
+		std::atomic_bool stop {false}; // disconnect
+		std::atomic_bool finished {false}; // disconnect
+
+		// pump thread
+		std::thread pump_thread;
+
+		// frame interval counters and estimates
+		// TODO
+
+		Connection(void) = default;
+		Connection(
+			ObjectHandle src_,
+			ObjectHandle sink_,
+			std::unique_ptr<Data>&& data_,
+			std::function<void(Connection&)>&& pump_fn_,
+			std::function<void(Connection&)>&& unsubscribe_fn_,
+			bool on_main_thread_ = true
+		);
+	};
+	std::vector<std::unique_ptr<Connection>> _connections;
+
+	std::unordered_map<std::string, Object> _default_sources;
+	std::unordered_map<std::string, Object> _default_sinks;
+
+	public:
+		StreamManager(ObjectStore2& os);
+		virtual ~StreamManager(void);
+
+		template<typename FrameType>
+		bool connect(Object src, Object sink, bool threaded = true);
+
+		bool connect(Object src, Object sink, bool threaded = true);
+		bool disconnect(Object src, Object sink);
+		bool disconnectAll(Object o);
+
+		// do we need the time delta?
+		float tick(float);
+
+	protected:
+		bool onEvent(const ObjectStore::Events::ObjectConstruct&) override;
+		bool onEvent(const ObjectStore::Events::ObjectUpdate&) override;
+		bool onEvent(const ObjectStore::Events::ObjectDestory&) override;
+};
+
+// template impls
+
+namespace Components {
+
+	// we require the complete sm type here
+	template<typename FrameType>
+	StreamSource StreamSource::create(const std::string& name) {
+		return StreamSource{
+			name,
+			std::string{entt::type_name<FrameType>::value()},
+			+[](StreamManager& sm, Object src, Object sink, bool threaded) {
+				return sm.connect<FrameType>(src, sink, threaded);
+			},
+		};
+	}
+
+	template<typename FrameType>
+	StreamSink StreamSink::create(const std::string& name) {
+		return StreamSink{
+			name,
+			std::string{entt::type_name<FrameType>::value()},
+		};
+	}
+
+} // Components
+
+template<typename FrameType>
+bool StreamManager::connect(Object src, Object sink, bool threaded) {
+	auto res = std::find_if(
+		_connections.cbegin(), _connections.cend(),
+		[&](const auto& a) { return a->src == src && a->sink == sink; }
+	);
+	if (res != _connections.cend()) {
+		// already exists
+		return false;
+	}
+
+	auto h_src = _os.objectHandle(src);
+	auto h_sink = _os.objectHandle(sink);
+	if (!static_cast<bool>(h_src) || !static_cast<bool>(h_sink)) {
+		// an object does not exist
+		return false;
+	}
+
+	if (!h_src.all_of<Components::FrameStream2Source<FrameType>>()) {
+		// src not stream source
+		return false;
+	}
+
+	if (!h_sink.all_of<Components::FrameStream2Sink<FrameType>>()) {
+		// sink not stream sink
+		return false;
+	}
+
+	auto& src_stream = h_src.get<Components::FrameStream2Source<FrameType>>();
+	auto& sink_stream = h_sink.get<Components::FrameStream2Sink<FrameType>>();
+
+	struct inlineData : public Connection::Data {
+		virtual ~inlineData(void) {}
+		std::shared_ptr<FrameStream2I<FrameType>> reader;
+		std::shared_ptr<FrameStream2I<FrameType>> writer;
+	};
+
+	auto our_data = std::make_unique<inlineData>();
+
+	our_data->reader = src_stream->subscribe();
+	if (!our_data->reader) {
+		return false;
+	}
+	our_data->writer = sink_stream->subscribe();
+	if (!our_data->writer) {
+		return false;
+	}
+
+	_connections.push_back(std::make_unique<Connection>(
+		h_src,
+		h_sink,
+		std::move(our_data),
+		[](Connection& con) -> void {
+			// there might be more stored
+			for (size_t i = 0; i < 10; i++) {
+				auto new_frame_opt = static_cast<inlineData*>(con.data.get())->reader->pop();
+				// TODO: frame interval estimates
+				if (new_frame_opt.has_value()) {
+					static_cast<inlineData*>(con.data.get())->writer->push(new_frame_opt.value());
+				} else {
+					break;
+				}
+			}
+		},
+		[](Connection& con) -> void {
+			auto* src_stream_ptr = con.src.try_get<Components::FrameStream2Source<FrameType>>();
+			if (src_stream_ptr != nullptr) {
+				(*src_stream_ptr)->unsubscribe(static_cast<inlineData*>(con.data.get())->reader);
+			}
+			auto* sink_stream_ptr = con.sink.try_get<Components::FrameStream2Sink<FrameType>>();
+			if (sink_stream_ptr != nullptr) {
+				(*sink_stream_ptr)->unsubscribe(static_cast<inlineData*>(con.data.get())->writer);
+			}
+		},
+		!threaded
+	));
+
+	return true;
+}
+

src/frame_streams/test_pop_reframer.cpp

@@ -0,0 +1,155 @@
+#include <iostream>
+
+#include "./audio_stream_pop_reframer.hpp"
+#include "./locked_frame_stream.hpp"
+
+#include <cassert>
+
+int main(void) {
+	{ // pump perfect
+		AudioStreamPopReFramer<LockedFrameStream2<AudioFrame2>> stream;
+		stream._frame_length_ms = 10;
+
+		AudioFrame2 f1 {
+			48'000,
+			1,
+			{},
+		};
+		f1.buffer = std::vector<int16_t>(
+			// perfect size
+			stream._frame_length_ms * f1.sample_rate * f1.channels / 1000,
+			0
+		);
+
+		{ // fill with sequential value
+			int16_t seq = 0;
+			for (auto& v : std::get<std::vector<int16_t>>(f1.buffer)) {
+				v = seq++;
+			}
+		}
+
+		stream.push(f1);
+
+		auto ret_opt = stream.pop();
+		assert(ret_opt);
+
+		auto& ret = ret_opt.value();
+		assert(ret.sample_rate == f1.sample_rate);
+		assert(ret.channels == f1.channels);
+		assert(ret.getSpan().size == f1.getSpan().size);
+		{
+			int16_t seq = 0;
+			for (const auto v : ret.getSpan()) {
+				assert(v == seq++);
+			}
+		}
+	}
+
+	{ // pump half
+		AudioStreamPopReFramer<LockedFrameStream2<AudioFrame2>> stream;
+		stream._frame_length_ms = 10;
+
+		AudioFrame2 f1 {
+			48'000,
+			1,
+			{},
+		};
+		f1.buffer = std::vector<int16_t>(
+			// perfect size
+			(stream._frame_length_ms * f1.sample_rate * f1.channels / 1000) / 2,
+			0
+		);
+		AudioFrame2 f2 {
+			48'000,
+			1,
+			{},
+		};
+		f2.buffer = std::vector<int16_t>(
+			// perfect size
+			(stream._frame_length_ms * f1.sample_rate * f1.channels / 1000) / 2,
+			0
+		);
+
+		{ // fill with sequential value
+			int16_t seq = 0;
+			for (auto& v : std::get<std::vector<int16_t>>(f1.buffer)) {
+				v = seq++;
+			}
+			for (auto& v : std::get<std::vector<int16_t>>(f2.buffer)) {
+				v = seq++;
+			}
+		}
+
+		stream.push(f1);
+		stream.push(f2);
+
+		// supposed to combine both
+		auto ret_opt = stream.pop();
+		assert(ret_opt);
+
+		auto& ret = ret_opt.value();
+		assert(ret.sample_rate == f1.sample_rate);
+		assert(ret.channels == f1.channels);
+		assert(ret.getSpan().size == stream._frame_length_ms * f1.sample_rate * f1.channels / 1000);
+		{
+			int16_t seq = 0;
+			for (const auto v : ret.getSpan()) {
+				assert(v == seq++);
+			}
+		}
+	}
+
+	{ // pump double
+		AudioStreamPopReFramer<LockedFrameStream2<AudioFrame2>> stream;
+		stream._frame_length_ms = 20;
+
+		AudioFrame2 f1 {
+			48'000,
+			2,
+			{},
+		};
+		f1.buffer = std::vector<int16_t>(
+			// perfect size
+			(stream._frame_length_ms * f1.sample_rate * f1.channels / 1000) * 2,
+			0
+		);
+		{ // fill with sequential value
+			int16_t seq = 0;
+			for (auto& v : std::get<std::vector<int16_t>>(f1.buffer)) {
+				v = seq++;
+			}
+		}
+
+		stream.push(f1);
+
+		// pop 2x
+		int16_t seq = 0;
+		{
+			auto ret_opt = stream.pop();
+			assert(ret_opt);
+
+			auto& ret = ret_opt.value();
+			assert(ret.sample_rate == f1.sample_rate);
+			assert(ret.channels == f1.channels);
+			assert(ret.getSpan().size == stream._frame_length_ms * f1.sample_rate * f1.channels / 1000);
+			for (const auto v : ret.getSpan()) {
+				assert(v == seq++);
+			}
+		}
+
+		{
+			auto ret_opt = stream.pop();
+			assert(ret_opt);
+
+			auto& ret = ret_opt.value();
+			assert(ret.sample_rate == f1.sample_rate);
+			assert(ret.channels == f1.channels);
+			assert(ret.getSpan().size == stream._frame_length_ms * f1.sample_rate * f1.channels / 1000);
+			for (const auto v : ret.getSpan()) {
+				assert(v == seq++);
+			}
+		}
+	}
+
+	return 0;
+}

src/frame_streams/voip_model.hpp

@@ -0,0 +1,77 @@
+#pragma once
+
+#include <solanaceae/contact/contact_model3.hpp>
+#include <solanaceae/object_store/fwd.hpp>
+
+struct VoIPModelI;
+
+namespace Components::VoIP {
+
+	struct TagVoIPSession {};
+
+	// getting called or invited by
+	struct Incoming {
+		Contact3 c{entt::null};
+	};
+
+	struct DefaultConfig {
+		bool incoming_audio {true};
+		bool incoming_video {true};
+		bool outgoing_audio {true};
+		bool outgoing_video {true};
+	};
+
+	// to talk to the model handling this session
+	//struct VoIPModel {
+		//VoIPModelI* ptr {nullptr};
+	//};
+
+	struct SessionState {
+		// ????
+		// incoming
+		// outgoing
+		enum class State {
+			RINGING,
+			CONNECTED,
+		} state;
+	};
+
+	struct SessionContact {
+		Contact3 c{entt::null};
+	};
+
+	struct StreamSources {
+		// list of all stream sources originating from this VoIP session
+		std::vector<Object> streams;
+	};
+
+	struct StreamSinks {
+		// list of all stream sinks going to this VoIP session
+		std::vector<Object> streams;
+	};
+
+} // Components::VoIP
+
+// TODO: events? piggyback on objects?
+
+// stream model instead?? -> more generic than "just" audio and video?
+// or specialized like this
+// streams abstract type in a nice way
+struct VoIPModelI {
+	virtual ~VoIPModelI(void) {}
+
+	// enters a call/voicechat/videocall ???
+	// - contact
+	// - default stream sources/sinks ?
+	// - enable a/v ? -> done on connecting to sources
+	// returns object tieing together the VoIP session
+	virtual ObjectHandle enter(const Contact3 c, const Components::VoIP::DefaultConfig& defaults = {true, true, true, true}) { (void)c,(void)defaults; return {}; }
+
+	// accept/join an invite to a session
+	virtual bool accept(ObjectHandle session, const Components::VoIP::DefaultConfig& defaults = {true, true, true, true}) { (void)session,(void)defaults; return false; }
+
+	// leaves a call
+	// - VoIP session object
+	virtual bool leave(ObjectHandle session) { (void)session; return false; }
+};
+

src/main.cpp

@@ -28,6 +28,8 @@ int main(int argc, char** argv) {
 
 	runSysCheck();
 
+	SDL_SetAppMetadata("tomato", "0.0.0-wip", nullptr);
+
 #ifdef __ANDROID__
 	// change current working dir to internal storage
 	std::filesystem::current_path(SDL_GetAndroidInternalStoragePath());
@@ -35,7 +37,7 @@ int main(int argc, char** argv) {
 
 	// setup hints
 #ifndef __ANDROID__
-	if (SDL_SetHint(SDL_HINT_VIDEO_ALLOW_SCREENSAVER, "1") != SDL_TRUE) {
+	if (!SDL_SetHint(SDL_HINT_VIDEO_ALLOW_SCREENSAVER, "1")) {
 		std::cerr << "Failed to set '" << SDL_HINT_VIDEO_ALLOW_SCREENSAVER << "' to 1\n";
 	}
 #endif

src/main_screen.cpp

@@ -5,6 +5,8 @@
 
 #include <solanaceae/contact/components.hpp>
 
+#include "./frame_streams/sdl/sdl_audio2_frame_stream2.hpp"
+
 #include <imgui/imgui.h>
 
 #include <SDL3/SDL.h>
@@ -19,16 +21,18 @@ MainScreen::MainScreen(SimpleConfigModel&& conf_, SDL_Renderer* renderer_, Theme
 	rmm(cr),
 	msnj{cr, {}, {}},
 	mts(rmm),
+	sm(os),
 	tc(save_path, save_password),
 	tpi(tc.getTox()),
 	ad(tc),
-#if TOMATO_TOX_AV
-	tav(tc.getTox()),
-#endif
 	tcm(cr, tc, tc),
 	tmm(rmm, cr, tcm, tc, tc),
 	ttm(rmm, cr, tcm, tc, tc, os),
 	tffom(cr, rmm, tcm, tc, tc),
+#if TOMATO_TOX_AV
+	tav(tc.getTox()),
+	tavvoip(os, tav, cr, tcm),
+#endif
 	theme(theme_),
 	mmil(rmm),
 	tam(/*rmm, */ os, cr, conf),
@@ -41,7 +45,9 @@ MainScreen::MainScreen(SimpleConfigModel&& conf_, SDL_Renderer* renderer_, Theme
 	sw(conf),
 	osui(os),
 	tuiu(tc, conf),
-	tdch(tpi)
+	tdch(tpi),
+	smui(os, sm),
+	dvt(os, sm, sdlrtu)
 {
 	tel.subscribeAll(tc);
 
@@ -75,7 +81,7 @@ MainScreen::MainScreen(SimpleConfigModel&& conf_, SDL_Renderer* renderer_, Theme
 		g_provideInstance<ToxPrivateI>("ToxPrivateI", "host", &tpi);
 		g_provideInstance<ToxEventProviderI>("ToxEventProviderI", "host", &tc);
 #if TOMATO_TOX_AV
-		g_provideInstance<ToxAV>("ToxAV", "host", &tav);
+		g_provideInstance<ToxAVI>("ToxAVI", "host", &tav);
 #endif
 		g_provideInstance<ToxContactModel2>("ToxContactModel2", "host", &tcm);
 
@@ -136,9 +142,46 @@ MainScreen::MainScreen(SimpleConfigModel&& conf_, SDL_Renderer* renderer_, Theme
 	}
 
 	conf.dump();
+
+	if (SDL_InitSubSystem(SDL_INIT_AUDIO)) {
+		// add system audio devices
+		{ // audio in
+			ObjectHandle asrc {os.registry(), os.registry().create()};
+			try {
+				asrc.emplace<Components::FrameStream2Source<AudioFrame2>>(
+					std::make_unique<SDLAudio2InputDevice>()
+				);
+
+				asrc.emplace<Components::StreamSource>(Components::StreamSource::create<AudioFrame2>("SDL Audio Default Recording Device"));
+				asrc.emplace<Components::TagDefaultTarget>();
+
+				os.throwEventConstruct(asrc);
+			} catch (...) {
+				os.registry().destroy(asrc);
+			}
+		}
+		{ // audio out
+			ObjectHandle asink {os.registry(), os.registry().create()};
+			try {
+				asink.emplace<Components::FrameStream2Sink<AudioFrame2>>(
+					std::make_unique<SDLAudio2OutputDeviceDefaultSink>()
+				);
+
+				asink.emplace<Components::StreamSink>(Components::StreamSink::create<AudioFrame2>("SDL Audio Default Playback Device"));
+				asink.emplace<Components::TagDefaultTarget>();
+
+				os.throwEventConstruct(asink);
+			} catch (...) {
+				os.registry().destroy(asink);
+			}
+		}
+	} else {
+		std::cerr << "MS warning: no sdl audio: " << SDL_GetError() << "\n";
+	}
 }
 
 MainScreen::~MainScreen(void) {
+	// TODO: quit sdl audio
 }
 
 bool MainScreen::handleEvent(SDL_Event& e) {
@@ -260,6 +303,8 @@ Screen* MainScreen::render(float time_delta, bool&) {
 	osui.render();
 	tuiu.render(); // render
 	tdch.render(); // render
+	smui.render();
+	const float dvt_interval = dvt.render();
 
 	{ // main window menubar injection
 		if (ImGui::Begin("tomato")) {
@@ -442,6 +487,7 @@ Screen* MainScreen::render(float time_delta, bool&) {
 	if (!_window_hidden && _time_since_event < curr_profile.low_delay_window) {
 		_render_interval = std::min<float>(_render_interval, ctc_interval);
 		_render_interval = std::min<float>(_render_interval, msgtc_interval);
+		_render_interval = std::min<float>(_render_interval, dvt_interval);
 
 		_render_interval = std::clamp(
 			_render_interval,
@@ -452,6 +498,7 @@ Screen* MainScreen::render(float time_delta, bool&) {
 	} else if (!_window_hidden && _time_since_event < curr_profile.mid_delay_window) {
 		_render_interval = std::min<float>(_render_interval, ctc_interval);
 		_render_interval = std::min<float>(_render_interval, msgtc_interval);
+		_render_interval = std::min<float>(_render_interval, dvt_interval);
 
 		_render_interval = std::clamp(
 			_render_interval,
@@ -474,8 +521,20 @@ Screen* MainScreen::render(float time_delta, bool&) {
 }
 
 Screen* MainScreen::tick(float time_delta, bool& quit) {
+	const float sm_interval = sm.tick(time_delta);
+
 	quit = !tc.iterate(time_delta); // compute
 
+#if TOMATO_TOX_AV
+	tav.toxavIterate();
+	// breaks it
+	// HACK: pow by 1.18 to increase 200 -> ~500
+	//const float av_interval = std::pow(tav.toxavIterationInterval(), 1.18)/1000.f;
+	const float av_interval = tav.toxavIterationInterval()/1000.f;
+
+	tavvoip.tick();
+#endif
+
 	tcm.iterate(time_delta); // compute
 
 	const float fo_interval = tffom.tick(time_delta);
@@ -505,11 +564,22 @@ Screen* MainScreen::tick(float time_delta, bool& quit) {
 		std::pow(tc.toxIterationInterval(), 1.6f)/1000.f,
 		pm_interval
 	);
+	_min_tick_interval = std::min<float>(
+		_min_tick_interval,
+		sm_interval
+	);
 	_min_tick_interval = std::min<float>(
 		_min_tick_interval,
 		fo_interval
 	);
 
+#if TOMATO_TOX_AV
+	_min_tick_interval = std::min<float>(
+		_min_tick_interval,
+		av_interval
+	);
+#endif
+
 	//std::cout << "MS: min tick interval: " << _min_tick_interval << "\n";
 
 	switch (_compute_perf_mode) {

src/main_screen.hpp

@@ -11,6 +11,7 @@
 #include <solanaceae/plugin/plugin_manager.hpp>
 #include <solanaceae/toxcore/tox_event_logger.hpp>
 #include "./tox_private_impl.hpp"
+#include "./frame_streams/stream_manager.hpp"
 
 #include <solanaceae/tox_contacts/tox_contact_model2.hpp>
 #include <solanaceae/tox_messages/tox_message_manager.hpp>
@@ -33,9 +34,12 @@
 #include "./tox_ui_utils.hpp"
 #include "./tox_dht_cap_histo.hpp"
 #include "./tox_friend_faux_offline_messaging.hpp"
+#include "./stream_manager_ui.hpp"
+#include "./debug_video_tap.hpp"
 
 #if TOMATO_TOX_AV
 #include "./tox_av.hpp"
+#include "./tox_av_voip_model.hpp"
 #endif
 
 #include <string>
@@ -58,17 +62,20 @@ struct MainScreen final : public Screen {
 	MessageSerializerNJ msnj;
 	MessageTimeSort mts;
 
+	StreamManager sm;
+
 	ToxEventLogger tel{std::cout};
 	ToxClient tc;
 	ToxPrivateImpl tpi;
 	AutoDirty ad;
-#if TOMATO_TOX_AV
-	ToxAV tav;
-#endif
 	ToxContactModel2 tcm;
 	ToxMessageManager tmm;
 	ToxTransferManager ttm;
 	ToxFriendFauxOfflineMessaging tffom;
+#if TOMATO_TOX_AV
+	ToxAVI tav;
+	ToxAVVoIPModel tavvoip;
+#endif
 
 	Theme& theme;
 
@@ -88,6 +95,8 @@ struct MainScreen final : public Screen {
 	ObjectStoreUI osui;
 	ToxUIUtils tuiu;
 	ToxDHTCapHisto tdch;
+	StreamManagerUI smui;
+	DebugVideoTap dvt;
 
 	PluginManager pm; // last, so it gets destroyed first
 

src/sdlrenderer_texture_uploader.cpp

@@ -32,7 +32,7 @@ uint64_t SDLRendererTextureUploader::uploadRGBA(const uint8_t* data, uint32_t wi
 	SDL_UpdateTexture(tex, nullptr, surf->pixels, surf->pitch);
 
 	SDL_BlendMode surf_blend_mode = SDL_BLENDMODE_NONE;
-	if (SDL_GetSurfaceBlendMode(surf, &surf_blend_mode) == 0) {
+	if (SDL_GetSurfaceBlendMode(surf, &surf_blend_mode)) {
 		SDL_SetTextureBlendMode(tex, surf_blend_mode);
 	}
 

src/stream_manager_ui.cpp

@@ -0,0 +1,234 @@
+#include "./stream_manager_ui.hpp"
+
+#include <solanaceae/object_store/object_store.hpp>
+
+#include <imgui/imgui.h>
+
+#include <string>
+
+StreamManagerUI::StreamManagerUI(ObjectStore2& os, StreamManager& sm) : _os(os), _sm(sm) {
+}
+
+void StreamManagerUI::render(void) {
+	{ // main window menubar injection
+		// assumes the window "tomato" was rendered already by cg
+		if (ImGui::Begin("tomato")) {
+			if (ImGui::BeginMenuBar()) {
+				// TODO: drop all menu sep?
+				//ImGui::Separator(); // os already exists (very hacky)
+				if (ImGui::BeginMenu("ObjectStore")) {
+					if (ImGui::MenuItem("Stream Manger", nullptr, _show_window)) {
+						_show_window = !_show_window;
+					}
+					ImGui::EndMenu();
+				}
+				ImGui::EndMenuBar();
+			}
+
+		}
+		ImGui::End();
+	}
+
+
+	if (!_show_window) {
+		return;
+	}
+
+	if (ImGui::Begin("StreamManagerUI", &_show_window)) {
+		// TODO: node canvas
+
+		// by fametype ??
+
+		if (ImGui::CollapsingHeader("Sources", ImGuiTreeNodeFlags_DefaultOpen)) {
+			// list sources
+			if (ImGui::BeginTable("sources_and_sinks", 4, ImGuiTableFlags_SizingFixedFit | ImGuiTableFlags_BordersInnerV)) {
+				ImGui::TableSetupColumn("id");
+				ImGui::TableSetupColumn("name");
+				ImGui::TableSetupColumn("##conn");
+				ImGui::TableSetupColumn("type");
+
+				ImGui::TableHeadersRow();
+
+				for (const auto& [oc, ss] : _os.registry().view<Components::StreamSource>().each()) {
+					//ImGui::Text("src  %d (%s)[%s]", entt::to_integral(entt::to_entity(oc)), ss.name.c_str(), ss.frame_type_name.c_str());
+					ImGui::PushID(entt::to_integral(oc));
+
+					ImGui::TableNextColumn();
+					ImGui::Text("%d", entt::to_integral(entt::to_entity(oc)));
+
+					if (_os.registry().all_of<Components::TagDefaultTarget>(oc)) {
+						ImGui::TableSetBgColor(ImGuiTableBgTarget_RowBg1, ImGui::GetColorU32(ImVec4{0.6f, 0.f, 0.6f, 0.25f}));
+					} else if (_os.registry().all_of<Components::TagConnectToDefault>(oc)) {
+						ImGui::TableSetBgColor(ImGuiTableBgTarget_RowBg1, ImGui::GetColorU32(ImVec4{0.6f, 0.6f, 0.f, 0.25f}));
+					}
+
+					const auto *ssrc = _os.registry().try_get<Components::StreamSource>(oc);
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(ssrc!=nullptr?ssrc->name.c_str():"none");
+
+					ImGui::TableNextColumn();
+					if (ImGui::SmallButton("->")) {
+						ImGui::OpenPopup("src_connect");
+					}
+					if (ImGui::BeginPopup("src_connect")) {
+						if (ImGui::BeginMenu("connect to")) {
+							for (const auto& [oc_sink, s_sink] : _os.registry().view<Components::StreamSink>().each()) {
+								if (s_sink.frame_type_name != ss.frame_type_name) {
+									continue;
+								}
+
+								ImGui::PushID(entt::to_integral(oc_sink));
+
+								std::string sink_label {"src "};
+								sink_label += std::to_string(entt::to_integral(entt::to_entity(oc_sink)));
+								sink_label += " (";
+								sink_label += s_sink.name;
+								sink_label += ")[";
+								sink_label += s_sink.frame_type_name;
+								sink_label += "]";
+								if (ImGui::MenuItem(sink_label.c_str())) {
+									_sm.connect(oc, oc_sink);
+								}
+
+								ImGui::PopID();
+							}
+
+							ImGui::EndMenu();
+						}
+						ImGui::EndPopup();
+					}
+
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(ssrc!=nullptr?ssrc->frame_type_name.c_str():"???");
+
+					ImGui::PopID();
+				}
+
+				ImGui::EndTable();
+			}
+		} // sources header
+
+		if (ImGui::CollapsingHeader("Sinks", ImGuiTreeNodeFlags_DefaultOpen)) {
+			// list sinks
+			if (ImGui::BeginTable("sources_and_sinks", 4, ImGuiTableFlags_SizingFixedFit | ImGuiTableFlags_BordersInnerV)) {
+				ImGui::TableSetupColumn("id");
+				ImGui::TableSetupColumn("name");
+				ImGui::TableSetupColumn("##conn");
+				ImGui::TableSetupColumn("type");
+
+				ImGui::TableHeadersRow();
+
+				for (const auto& [oc, ss] : _os.registry().view<Components::StreamSink>().each()) {
+					ImGui::PushID(entt::to_integral(oc));
+
+					ImGui::TableNextColumn();
+					ImGui::Text("%d", entt::to_integral(entt::to_entity(oc)));
+
+					if (_os.registry().all_of<Components::TagDefaultTarget>(oc)) {
+						ImGui::TableSetBgColor(ImGuiTableBgTarget_RowBg1, ImGui::GetColorU32(ImVec4{0.6f, 0.f, 0.6f, 0.25f}));
+					} else if (_os.registry().all_of<Components::TagConnectToDefault>(oc)) {
+						ImGui::TableSetBgColor(ImGuiTableBgTarget_RowBg1, ImGui::GetColorU32(ImVec4{0.6f, 0.6f, 0.f, 0.25f}));
+					}
+
+					const auto *ssink = _os.registry().try_get<Components::StreamSink>(oc);
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(ssink!=nullptr?ssink->name.c_str():"none");
+
+					ImGui::TableNextColumn();
+					if (ImGui::SmallButton("->")) {
+						ImGui::OpenPopup("sink_connect");
+					}
+					// ImGuiWindowFlags_AlwaysAutoResize | ImGuiWindowFlags_NoTitleBar | ImGuiWindowFlags_NoSavedSettings
+					if (ImGui::BeginPopup("sink_connect")) {
+						if (ImGui::BeginMenu("connect to")) {
+							for (const auto& [oc_src, s_src] : _os.registry().view<Components::StreamSource>().each()) {
+								if (s_src.frame_type_name != ss.frame_type_name) {
+									continue;
+								}
+
+								ImGui::PushID(entt::to_integral(oc_src));
+
+								std::string source_label {"src "};
+								source_label += std::to_string(entt::to_integral(entt::to_entity(oc_src)));
+								source_label += " (";
+								source_label += s_src.name;
+								source_label += ")[";
+								source_label += s_src.frame_type_name;
+								source_label += "]";
+								if (ImGui::MenuItem(source_label.c_str())) {
+									_sm.connect(oc_src, oc);
+								}
+
+								ImGui::PopID();
+							}
+
+							ImGui::EndMenu();
+						}
+						ImGui::EndPopup();
+					}
+
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(ssink!=nullptr?ssink->frame_type_name.c_str():"???");
+
+					ImGui::PopID();
+				}
+
+				ImGui::EndTable();
+			}
+		} // sink header
+
+		if (ImGui::CollapsingHeader("Connections", ImGuiTreeNodeFlags_DefaultOpen)) {
+			// list connections
+			if (ImGui::BeginTable("connections", 6, ImGuiTableFlags_SizingFixedFit | ImGuiTableFlags_BordersInnerV)) {
+				ImGui::TableSetupColumn("##id"); // TODO: remove?
+				ImGui::TableSetupColumn("##disco");
+				ImGui::TableSetupColumn("##qdesc");
+				ImGui::TableSetupColumn("from");
+				ImGui::TableSetupColumn("to");
+				ImGui::TableSetupColumn("type");
+
+				ImGui::TableHeadersRow();
+
+				for (size_t i = 0; i < _sm._connections.size(); i++) {
+					const auto& con = _sm._connections[i];
+					//ImGui::Text("con %d->%d", entt::to_integral(entt::to_entity(con->src.entity())), entt::to_integral(entt::to_entity(con->sink.entity())));
+
+					ImGui::PushID(i);
+
+					ImGui::TableNextColumn();
+					ImGui::Text("%zu", i); // do connections have ids?
+
+					ImGui::TableNextColumn();
+					if (ImGui::SmallButton("X")) {
+						con->stop = true;
+					}
+
+					ImGui::TableNextColumn();
+					ImGui::Text("%d->%d", entt::to_integral(entt::to_entity(con->src.entity())), entt::to_integral(entt::to_entity(con->sink.entity())));
+
+					const auto *ssrc = con->src.try_get<Components::StreamSource>();
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(ssrc!=nullptr?ssrc->name.c_str():"none");
+
+					const auto *ssink = con->sink.try_get<Components::StreamSink>();
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(ssink!=nullptr?ssink->name.c_str():"none");
+
+					ImGui::TableNextColumn();
+					ImGui::TextUnformatted(
+						(ssrc!=nullptr)?
+							ssrc->frame_type_name.c_str():
+							(ssink!=nullptr)?
+								ssink->frame_type_name.c_str()
+								:"???"
+					);
+
+					ImGui::PopID();
+				}
+				ImGui::EndTable();
+			}
+		} // con header
+	}
+	ImGui::End();
+}
+

src/stream_manager_ui.hpp

@@ -0,0 +1,17 @@
+#pragma once
+
+#include <solanaceae/object_store/fwd.hpp>
+#include "./frame_streams/stream_manager.hpp"
+
+class StreamManagerUI {
+	ObjectStore2& _os;
+	StreamManager& _sm;
+
+	bool _show_window {false};
+
+	public:
+		StreamManagerUI(ObjectStore2& os, StreamManager& sm);
+
+		void render(void);
+};
+

src/tox_av.cpp

@@ -2,81 +2,239 @@
 
 #include <cassert>
 
+#include <cstdint>
+#include <iostream>
+
 // https://almogfx.bandcamp.com/track/crushed-w-cassade
 
-ToxAV::ToxAV(Tox* tox) : _tox(tox) {
+ToxAVI::ToxAVI(Tox* tox) : _tox(tox) {
 	Toxav_Err_New err_new {TOXAV_ERR_NEW_OK};
 	_tox_av = toxav_new(_tox, &err_new);
 	// TODO: throw
 	assert(err_new == TOXAV_ERR_NEW_OK);
+
+	toxav_callback_call(
+		_tox_av,
+		+[](ToxAV*, uint32_t friend_number, bool audio_enabled, bool video_enabled, void *user_data) {
+			assert(user_data != nullptr);
+			static_cast<ToxAVI*>(user_data)->cb_call(friend_number, audio_enabled, video_enabled);
+		},
+		this
+	);
+	toxav_callback_call_state(
+		_tox_av,
+		+[](ToxAV*, uint32_t friend_number, uint32_t state, void *user_data) {
+			assert(user_data != nullptr);
+			static_cast<ToxAVI*>(user_data)->cb_call_state(friend_number, state);
+		},
+		this
+	);
+	toxav_callback_audio_bit_rate(
+		_tox_av,
+		+[](ToxAV*, uint32_t friend_number, uint32_t audio_bit_rate, void *user_data) {
+			assert(user_data != nullptr);
+			static_cast<ToxAVI*>(user_data)->cb_audio_bit_rate(friend_number, audio_bit_rate);
+		},
+		this
+	);
+	toxav_callback_video_bit_rate(
+		_tox_av,
+		+[](ToxAV*, uint32_t friend_number, uint32_t video_bit_rate, void *user_data) {
+			assert(user_data != nullptr);
+			static_cast<ToxAVI*>(user_data)->cb_video_bit_rate(friend_number, video_bit_rate);
+		},
+		this
+	);
+	toxav_callback_audio_receive_frame(
+		_tox_av,
+		+[](ToxAV*, uint32_t friend_number, const int16_t pcm[], size_t sample_count, uint8_t channels, uint32_t sampling_rate, void *user_data) {
+			assert(user_data != nullptr);
+			static_cast<ToxAVI*>(user_data)->cb_audio_receive_frame(friend_number, pcm, sample_count, channels, sampling_rate);
+		},
+		this
+	);
+	toxav_callback_video_receive_frame(
+		_tox_av,
+		+[](ToxAV*, uint32_t friend_number,
+			uint16_t width, uint16_t height,
+			const uint8_t y[/*! max(width, abs(ystride)) * height */],
+			const uint8_t u[/*! max(width/2, abs(ustride)) * (height/2) */],
+			const uint8_t v[/*! max(width/2, abs(vstride)) * (height/2) */],
+			int32_t ystride, int32_t ustride, int32_t vstride,
+			void *user_data
+		) {
+			assert(user_data != nullptr);
+			static_cast<ToxAVI*>(user_data)->cb_video_receive_frame(friend_number, width, height, y, u, v, ystride, ustride, vstride);
+		},
+		this
+	);
 }
-ToxAV::~ToxAV(void) {
+
+ToxAVI::~ToxAVI(void) {
 	toxav_kill(_tox_av);
 }
 
-uint32_t ToxAV::toxavIterationInterval(void) const {
+uint32_t ToxAVI::toxavIterationInterval(void) const {
 	return toxav_iteration_interval(_tox_av);
 }
 
-void ToxAV::toxavIterate(void) {
+void ToxAVI::toxavIterate(void) {
 	toxav_iterate(_tox_av);
 }
 
-uint32_t ToxAV::toxavAudioIterationInterval(void) const {
+uint32_t ToxAVI::toxavAudioIterationInterval(void) const {
 	return toxav_audio_iteration_interval(_tox_av);
 }
 
-void ToxAV::toxavAudioIterate(void) {
+void ToxAVI::toxavAudioIterate(void) {
 	toxav_audio_iterate(_tox_av);
 }
 
-uint32_t ToxAV::toxavVideoIterationInterval(void) const {
+uint32_t ToxAVI::toxavVideoIterationInterval(void) const {
 	return toxav_video_iteration_interval(_tox_av);
 }
 
-void ToxAV::toxavVideoIterate(void) {
+void ToxAVI::toxavVideoIterate(void) {
 	toxav_video_iterate(_tox_av);
 }
 
-Toxav_Err_Call ToxAV::toxavCall(uint32_t friend_number, uint32_t audio_bit_rate, uint32_t video_bit_rate) {
+Toxav_Err_Call ToxAVI::toxavCall(uint32_t friend_number, uint32_t audio_bit_rate, uint32_t video_bit_rate) {
 	Toxav_Err_Call err {TOXAV_ERR_CALL_OK};
 	toxav_call(_tox_av, friend_number, audio_bit_rate, video_bit_rate, &err);
 	return err;
 }
 
-Toxav_Err_Answer ToxAV::toxavAnswer(uint32_t friend_number, uint32_t audio_bit_rate, uint32_t video_bit_rate) {
+Toxav_Err_Answer ToxAVI::toxavAnswer(uint32_t friend_number, uint32_t audio_bit_rate, uint32_t video_bit_rate) {
 	Toxav_Err_Answer err {TOXAV_ERR_ANSWER_OK};
 	toxav_answer(_tox_av, friend_number, audio_bit_rate, video_bit_rate, &err);
 	return err;
 }
 
-Toxav_Err_Call_Control ToxAV::toxavCallControl(uint32_t friend_number, Toxav_Call_Control control) {
+Toxav_Err_Call_Control ToxAVI::toxavCallControl(uint32_t friend_number, Toxav_Call_Control control) {
 	Toxav_Err_Call_Control err {TOXAV_ERR_CALL_CONTROL_OK};
 	toxav_call_control(_tox_av, friend_number, control, &err);
 	return err;
 }
 
-Toxav_Err_Send_Frame ToxAV::toxavAudioSendFrame(uint32_t friend_number, const int16_t pcm[], size_t sample_count, uint8_t channels, uint32_t sampling_rate) {
+Toxav_Err_Send_Frame ToxAVI::toxavAudioSendFrame(uint32_t friend_number, const int16_t pcm[], size_t sample_count, uint8_t channels, uint32_t sampling_rate) {
 	Toxav_Err_Send_Frame err {TOXAV_ERR_SEND_FRAME_OK};
 	toxav_audio_send_frame(_tox_av, friend_number, pcm, sample_count, channels, sampling_rate, &err);
 	return err;
 }
 
-Toxav_Err_Bit_Rate_Set ToxAV::toxavAudioSetBitRate(uint32_t friend_number, uint32_t bit_rate) {
+Toxav_Err_Bit_Rate_Set ToxAVI::toxavAudioSetBitRate(uint32_t friend_number, uint32_t bit_rate) {
 	Toxav_Err_Bit_Rate_Set err {TOXAV_ERR_BIT_RATE_SET_OK};
 	toxav_audio_set_bit_rate(_tox_av, friend_number, bit_rate, &err);
 	return err;
 }
 
-Toxav_Err_Send_Frame ToxAV::toxavVideoSendFrame(uint32_t friend_number, uint16_t width, uint16_t height, const uint8_t y[], const uint8_t u[], const uint8_t v[]) {
+Toxav_Err_Send_Frame ToxAVI::toxavVideoSendFrame(uint32_t friend_number, uint16_t width, uint16_t height, const uint8_t y[], const uint8_t u[], const uint8_t v[]) {
 	Toxav_Err_Send_Frame err {TOXAV_ERR_SEND_FRAME_OK};
 	toxav_video_send_frame(_tox_av, friend_number, width, height, y, u, v, &err);
 	return err;
 }
 
-Toxav_Err_Bit_Rate_Set ToxAV::toxavVideoSetBitRate(uint32_t friend_number, uint32_t bit_rate) {
+Toxav_Err_Bit_Rate_Set ToxAVI::toxavVideoSetBitRate(uint32_t friend_number, uint32_t bit_rate) {
 	Toxav_Err_Bit_Rate_Set err {TOXAV_ERR_BIT_RATE_SET_OK};
 	toxav_video_set_bit_rate(_tox_av, friend_number, bit_rate, &err);
 	return err;
 }
 
+void ToxAVI::cb_call(uint32_t friend_number, bool audio_enabled, bool video_enabled) {
+	std::cerr << "TOXAV: receiving call f:" << friend_number << " a:" << audio_enabled << " v:" << video_enabled << "\n";
+	//Toxav_Err_Answer err_answer { TOXAV_ERR_ANSWER_OK };
+	//toxav_answer(_tox_av, friend_number, 0, 0, &err_answer);
+	//if (err_answer != TOXAV_ERR_ANSWER_OK) {
+	//    std::cerr << "!!!!!!!! answer failed " << err_answer << "\n";
+	//}
+
+	dispatch(
+		ToxAV_Event::friend_call,
+		Events::FriendCall{
+			friend_number,
+			audio_enabled,
+			video_enabled,
+		}
+	);
+}
+
+void ToxAVI::cb_call_state(uint32_t friend_number, uint32_t state) {
+	//ToxAVFriendCallState w_state{state};
+
+	//w_state.is_error();
+
+	std::cerr << "TOXAV: call state f:" << friend_number << " s:" << state << "\n";
+
+	dispatch(
+		ToxAV_Event::friend_call_state,
+		Events::FriendCallState{
+			friend_number,
+			state,
+		}
+	);
+}
+
+void ToxAVI::cb_audio_bit_rate(uint32_t friend_number, uint32_t audio_bit_rate) {
+	std::cerr << "TOXAV: audio bitrate f:" << friend_number << " abr:" << audio_bit_rate << "\n";
+
+	dispatch(
+		ToxAV_Event::friend_audio_bitrate,
+		Events::FriendAudioBitrate{
+			friend_number,
+			audio_bit_rate,
+		}
+	);
+}
+
+void ToxAVI::cb_video_bit_rate(uint32_t friend_number, uint32_t video_bit_rate) {
+	std::cerr << "TOXAV: video bitrate f:" << friend_number << " vbr:" << video_bit_rate << "\n";
+
+	dispatch(
+		ToxAV_Event::friend_video_bitrate,
+		Events::FriendVideoBitrate{
+			friend_number,
+			video_bit_rate,
+		}
+	);
+}
+
+void ToxAVI::cb_audio_receive_frame(uint32_t friend_number, const int16_t pcm[], size_t sample_count, uint8_t channels, uint32_t sampling_rate) {
+	//std::cerr << "TOXAV: audio frame f:" <<  friend_number << " sc:" << sample_count << " ch:" << (int)channels << " sr:" << sampling_rate << "\n";
+
+	dispatch(
+		ToxAV_Event::friend_audio_frame,
+		Events::FriendAudioFrame{
+			friend_number,
+			Span<int16_t>(pcm, sample_count*channels), // TODO: is sample count *ch or /ch?
+			channels,
+			sampling_rate,
+		}
+	);
+}
+
+void ToxAVI::cb_video_receive_frame(
+	uint32_t friend_number,
+	uint16_t width, uint16_t height,
+	const uint8_t y[/*! max(width, abs(ystride)) * height */],
+	const uint8_t u[/*! max(width/2, abs(ustride)) * (height/2) */],
+	const uint8_t v[/*! max(width/2, abs(vstride)) * (height/2) */],
+	int32_t ystride, int32_t ustride, int32_t vstride
+) {
+	//std::cerr << "TOXAV: video frame f:" <<  friend_number << " w:" << width << " h:" << height << "\n";
+
+	dispatch(
+		ToxAV_Event::friend_video_frame,
+		Events::FriendVideoFrame{
+			friend_number,
+			width,
+			height,
+			Span<uint8_t>(y, std::max<int64_t>(width, std::abs(ystride)) * height),
+			Span<uint8_t>(u, std::max<int64_t>(width/2, std::abs(ustride)) * (height/2)),
+			Span<uint8_t>(v, std::max<int64_t>(width/2, std::abs(vstride)) * (height/2)),
+			ystride,
+			ustride,
+			vstride,
+		}
+	);
+}
+

src/tox_av.hpp

@@ -1,15 +1,99 @@
 #pragma once
 
+#include <solanaceae/util/span.hpp>
+#include <solanaceae/util/event_provider.hpp>
+
 #include <tox/toxav.h>
 
-struct ToxAV {
+namespace /*toxav*/ Events {
+
+	struct FriendCall {
+		uint32_t friend_number;
+		bool audio_enabled;
+		bool video_enabled;
+	};
+
+	struct FriendCallState {
+		uint32_t friend_number;
+		uint32_t state;
+	};
+
+	struct FriendAudioBitrate {
+		uint32_t friend_number;
+		uint32_t audio_bit_rate;
+	};
+
+	struct FriendVideoBitrate {
+		uint32_t friend_number;
+		uint32_t video_bit_rate;
+	};
+
+	struct FriendAudioFrame {
+		uint32_t friend_number;
+
+		Span<int16_t> pcm;
+		//size_t sample_count;
+		uint8_t channels;
+		uint32_t sampling_rate;
+	};
+
+	struct FriendVideoFrame {
+		uint32_t friend_number;
+
+		uint16_t width;
+		uint16_t height;
+		//const uint8_t y[[>! max(width, abs(ystride)) * height <]];
+		//const uint8_t u[[>! max(width/2, abs(ustride)) * (height/2) <]];
+		//const uint8_t v[[>! max(width/2, abs(vstride)) * (height/2) <]];
+		// mdspan would be nice here
+		// bc of the stride, span might be larger than the actual data it contains
+		Span<uint8_t> y;
+		Span<uint8_t> u;
+		Span<uint8_t> v;
+		int32_t ystride;
+		int32_t ustride;
+		int32_t vstride;
+	};
+
+} // Event
+
+enum class ToxAV_Event : uint32_t {
+	friend_call,
+	friend_call_state,
+	friend_audio_bitrate,
+	friend_video_bitrate,
+	friend_audio_frame,
+	friend_video_frame,
+
+	MAX
+};
+
+struct ToxAVEventI {
+	using enumType = ToxAV_Event;
+
+	virtual ~ToxAVEventI(void) {}
+
+	virtual bool onEvent(const Events::FriendCall&) { return false; }
+	virtual bool onEvent(const Events::FriendCallState&) { return false; }
+	virtual bool onEvent(const Events::FriendAudioBitrate&) { return false; }
+	virtual bool onEvent(const Events::FriendVideoBitrate&) { return false; }
+	virtual bool onEvent(const Events::FriendAudioFrame&) { return false; }
+	virtual bool onEvent(const Events::FriendVideoFrame&) { return false; }
+};
+using ToxAVEventProviderI = EventProviderI<ToxAVEventI>;
+
+// TODO: seperate out implementation from interface
+struct ToxAVI : public ToxAVEventProviderI {
 	Tox* _tox = nullptr;
 	ToxAV* _tox_av = nullptr;
 
-	ToxAV(Tox* tox);
-	virtual ~ToxAV(void);
+	static constexpr const char* version {"0"};
+
+	ToxAVI(Tox* tox);
+	virtual ~ToxAVI(void);
 
 	// interface
+	// if iterate is called on a different thread, it will fire events there
 	uint32_t toxavIterationInterval(void) const;
 	void toxavIterate(void);
 
@@ -33,5 +117,32 @@ struct ToxAV {
 //int32_t toxav_groupchat_disable_av(Tox *tox, uint32_t groupnumber);
 //bool toxav_groupchat_av_enabled(Tox *tox, uint32_t groupnumber);
 
+
+
+	// toxav callbacks
+	void cb_call(uint32_t friend_number, bool audio_enabled, bool video_enabled);
+	void cb_call_state(uint32_t friend_number, uint32_t state);
+	void cb_audio_bit_rate(uint32_t friend_number, uint32_t audio_bit_rate);
+	void cb_video_bit_rate(uint32_t friend_number, uint32_t video_bit_rate);
+	void cb_audio_receive_frame(uint32_t friend_number, const int16_t pcm[], size_t sample_count, uint8_t channels, uint32_t sampling_rate);
+	void cb_video_receive_frame(
+		uint32_t friend_number,
+		uint16_t width, uint16_t height,
+		const uint8_t y[/*! max(width, abs(ystride)) * height */],
+		const uint8_t u[/*! max(width/2, abs(ustride)) * (height/2) */],
+		const uint8_t v[/*! max(width/2, abs(vstride)) * (height/2) */],
+		int32_t ystride, int32_t ustride, int32_t vstride
+	);
+};
+
+struct ToxAVFriendCallState final {
+	const uint32_t state {TOXAV_FRIEND_CALL_STATE_NONE};
+
+	[[nodiscard]] bool is_error(void) const { return state & TOXAV_FRIEND_CALL_STATE_ERROR; }
+	[[nodiscard]] bool is_finished(void) const { return state & TOXAV_FRIEND_CALL_STATE_FINISHED; }
+	[[nodiscard]] bool is_sending_a(void) const { return state & TOXAV_FRIEND_CALL_STATE_SENDING_A; }
+	[[nodiscard]] bool is_sending_v(void) const { return state & TOXAV_FRIEND_CALL_STATE_SENDING_V; }
+	[[nodiscard]] bool is_accepting_a(void) const { return state & TOXAV_FRIEND_CALL_STATE_ACCEPTING_A; }
+	[[nodiscard]] bool is_accepting_v(void) const { return state & TOXAV_FRIEND_CALL_STATE_ACCEPTING_V; }
 };
 

src/tox_av_voip_model.cpp

@@ -0,0 +1,711 @@
+#include "./tox_av_voip_model.hpp"
+
+#include <solanaceae/object_store/object_store.hpp>
+#include <solanaceae/tox_contacts/components.hpp>
+
+#include "./frame_streams/stream_manager.hpp"
+#include "./frame_streams/audio_stream2.hpp"
+#include "./frame_streams/locked_frame_stream.hpp"
+#include "./frame_streams/multi_source.hpp"
+#include "./frame_streams/audio_stream_pop_reframer.hpp"
+
+#include "./frame_streams/sdl/video.hpp"
+#include "./frame_streams/sdl/video_push_converter.hpp"
+
+#include <cstring>
+
+#include <iostream>
+
+// fwd
+namespace Message {
+uint64_t getTimeMS(void);
+} // Message
+
+namespace Components {
+	struct ToxAVIncomingAV {
+		bool incoming_audio {false};
+		bool incoming_video {false};
+	};
+
+	struct ToxAVAudioSink {
+		ObjectHandle o;
+		// ptr?
+	};
+	struct ToxAVVideoSink {
+		ObjectHandle o;
+	};
+	struct ToxAVAudioSource {
+		ObjectHandle o;
+	};
+	struct ToxAVVideoSource {
+		ObjectHandle o;
+	};
+} // Components
+
+struct ToxAVCallAudioSink : public FrameStream2SinkI<AudioFrame2> {
+	ToxAVI& _toxav;
+
+	// bitrate for enabled state
+	uint32_t _audio_bitrate {32};
+
+	uint32_t _fid;
+	std::shared_ptr<AudioStreamPopReFramer<LockedFrameStream2<AudioFrame2>>> _writer;
+
+	ToxAVCallAudioSink(ToxAVI& toxav, uint32_t fid) : _toxav(toxav), _fid(fid) {}
+	~ToxAVCallAudioSink(void) {
+		if (_writer) {
+			_writer = nullptr;
+			_toxav.toxavAudioSetBitRate(_fid, 0);
+		}
+	}
+
+	// sink
+	std::shared_ptr<FrameStream2I<AudioFrame2>> subscribe(void) override {
+		if (_writer) {
+			// max 1 (exclusive for now)
+			return nullptr;
+		}
+
+		auto err = _toxav.toxavAudioSetBitRate(_fid, _audio_bitrate);
+		if (err != TOXAV_ERR_BIT_RATE_SET_OK) {
+			return nullptr;
+		}
+
+		// 20ms for now, 10ms would work too, further investigate stutters at 5ms (probably too slow interval rate)
+		_writer = std::make_shared<AudioStreamPopReFramer<LockedFrameStream2<AudioFrame2>>>(20);
+
+		return _writer;
+	}
+
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<AudioFrame2>>& sub) override {
+		if (!sub || !_writer) {
+			// nah
+			return false;
+		}
+
+		if (sub == _writer) {
+			_writer = nullptr;
+
+			/*auto err = */_toxav.toxavAudioSetBitRate(_fid, 0);
+			// print warning? on error?
+
+			return true;
+		}
+
+		// what
+		return false;
+	}
+};
+
+// exlusive
+struct ToxAVCallVideoSink : public FrameStream2SinkI<SDLVideoFrame> {
+	using stream_type = PushConversionVideoStream<LockedFrameStream2<SDLVideoFrame>>;
+	ToxAVI& _toxav;
+
+	// bitrate for enabled state
+	uint32_t _video_bitrate {2}; // HACK: hardcode to 2mbits (toxap wrongly multiplies internally by 1000)
+
+	uint32_t _fid;
+	std::shared_ptr<stream_type> _writer;
+
+	ToxAVCallVideoSink(ToxAVI& toxav, uint32_t fid) : _toxav(toxav), _fid(fid) {}
+	~ToxAVCallVideoSink(void) {
+		if (_writer) {
+			_writer = nullptr;
+			_toxav.toxavVideoSetBitRate(_fid, 0);
+		}
+	}
+
+	// sink
+	std::shared_ptr<FrameStream2I<SDLVideoFrame>> subscribe(void) override {
+		if (_writer) {
+			// max 1 (exclusive, composite video somewhere else)
+			return nullptr;
+		}
+
+		auto err = _toxav.toxavVideoSetBitRate(_fid, _video_bitrate);
+		if (err != TOXAV_ERR_BIT_RATE_SET_OK) {
+			return nullptr;
+		}
+
+		// toxav needs I420
+		_writer = std::make_shared<stream_type>(SDL_PIXELFORMAT_IYUV);
+
+		return _writer;
+	}
+
+	bool unsubscribe(const std::shared_ptr<FrameStream2I<SDLVideoFrame>>& sub) override {
+		if (!sub || !_writer) {
+			// nah
+			return false;
+		}
+
+		if (sub == _writer) {
+			_writer = nullptr;
+
+			/*auto err = */_toxav.toxavVideoSetBitRate(_fid, 0);
+			// print warning? on error?
+
+			return true;
+		}
+
+		// what
+		return false;
+	}
+};
+
+void ToxAVVoIPModel::addAudioSource(ObjectHandle session, uint32_t friend_number) {
+	auto& stream_source = session.get_or_emplace<Components::VoIP::StreamSources>().streams;
+
+	ObjectHandle incoming_audio {_os.registry(), _os.registry().create()};
+
+	auto new_asrc = std::make_unique<FrameStream2MultiSource<AudioFrame2>>();
+	incoming_audio.emplace<FrameStream2MultiSource<AudioFrame2>*>(new_asrc.get());
+	incoming_audio.emplace<Components::FrameStream2Source<AudioFrame2>>(std::move(new_asrc));
+	incoming_audio.emplace<Components::StreamSource>(Components::StreamSource::create<AudioFrame2>("ToxAV Friend Call Incoming Audio"));
+
+	if (
+		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
+		defaults != nullptr && defaults->incoming_audio
+	) {
+		incoming_audio.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
+	}
+
+	stream_source.push_back(incoming_audio);
+	session.emplace<Components::ToxAVAudioSource>(incoming_audio);
+	// TODO: tie session to stream
+
+	_audio_sources[friend_number] = incoming_audio;
+
+	_os.throwEventConstruct(incoming_audio);
+}
+
+void ToxAVVoIPModel::addAudioSink(ObjectHandle session, uint32_t friend_number) {
+	auto& stream_sinks = session.get_or_emplace<Components::VoIP::StreamSinks>().streams;
+	ObjectHandle outgoing_audio {_os.registry(), _os.registry().create()};
+
+	auto new_asink = std::make_unique<ToxAVCallAudioSink>(_av, friend_number);
+	outgoing_audio.emplace<ToxAVCallAudioSink*>(new_asink.get());
+	outgoing_audio.emplace<Components::FrameStream2Sink<AudioFrame2>>(std::move(new_asink));
+	outgoing_audio.emplace<Components::StreamSink>(Components::StreamSink::create<AudioFrame2>("ToxAV Friend Call Outgoing Audio"));
+
+	if (
+		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
+		defaults != nullptr && defaults->outgoing_audio
+	) {
+		outgoing_audio.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
+	}
+
+	stream_sinks.push_back(outgoing_audio);
+	session.emplace<Components::ToxAVAudioSink>(outgoing_audio);
+	// TODO: tie session to stream
+
+	_os.throwEventConstruct(outgoing_audio);
+}
+
+void ToxAVVoIPModel::addVideoSource(ObjectHandle session, uint32_t friend_number) {
+	auto& stream_source = session.get_or_emplace<Components::VoIP::StreamSources>().streams;
+
+	ObjectHandle incoming_video {_os.registry(), _os.registry().create()};
+
+	auto new_vsrc = std::make_unique<FrameStream2MultiSource<SDLVideoFrame>>();
+	incoming_video.emplace<FrameStream2MultiSource<SDLVideoFrame>*>(new_vsrc.get());
+	incoming_video.emplace<Components::FrameStream2Source<SDLVideoFrame>>(std::move(new_vsrc));
+	incoming_video.emplace<Components::StreamSource>(Components::StreamSource::create<SDLVideoFrame>("ToxAV Friend Call Incoming Video"));
+
+	if (
+		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
+		defaults != nullptr && defaults->incoming_video
+	) {
+		incoming_video.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
+	}
+
+	stream_source.push_back(incoming_video);
+	session.emplace<Components::ToxAVVideoSource>(incoming_video);
+	// TODO: tie session to stream
+
+	_video_sources[friend_number] = incoming_video;
+
+	_os.throwEventConstruct(incoming_video);
+}
+
+void ToxAVVoIPModel::addVideoSink(ObjectHandle session, uint32_t friend_number) {
+	auto& stream_sinks = session.get_or_emplace<Components::VoIP::StreamSinks>().streams;
+	ObjectHandle outgoing_video {_os.registry(), _os.registry().create()};
+
+	auto new_vsink = std::make_unique<ToxAVCallVideoSink>(_av, friend_number);
+	outgoing_video.emplace<ToxAVCallVideoSink*>(new_vsink.get());
+	outgoing_video.emplace<Components::FrameStream2Sink<SDLVideoFrame>>(std::move(new_vsink));
+	outgoing_video.emplace<Components::StreamSink>(Components::StreamSink::create<SDLVideoFrame>("ToxAV Friend Call Outgoing Video"));
+
+	if (
+		const auto* defaults = session.try_get<Components::VoIP::DefaultConfig>();
+		defaults != nullptr && defaults->outgoing_video
+	) {
+		outgoing_video.emplace<Components::TagConnectToDefault>(); // depends on what was specified in enter()
+	}
+
+	stream_sinks.push_back(outgoing_video);
+	session.emplace<Components::ToxAVVideoSink>(outgoing_video);
+	// TODO: tie session to stream
+
+	_os.throwEventConstruct(outgoing_video);
+}
+
+void ToxAVVoIPModel::destroySession(ObjectHandle session) {
+	if (!static_cast<bool>(session)) {
+		return;
+	}
+
+	// remove lookup
+	if (session.all_of<Components::ToxAVAudioSource>()) {
+		auto it_asrc = std::find_if(
+			_audio_sources.cbegin(), _audio_sources.cend(),
+			[o = session.get<Components::ToxAVAudioSource>().o](const auto& it) {
+				return it.second == o;
+			}
+		);
+
+		if (it_asrc != _audio_sources.cend()) {
+			_audio_sources.erase(it_asrc);
+		}
+	}
+	if (session.all_of<Components::ToxAVVideoSource>()) {
+		auto it_vsrc = std::find_if(
+			_video_sources.cbegin(), _video_sources.cend(),
+			[o = session.get<Components::ToxAVVideoSource>().o](const auto& it) {
+				return it.second == o;
+			}
+		);
+
+		if (it_vsrc != _video_sources.cend()) {
+			_video_sources.erase(it_vsrc);
+		}
+	}
+
+	// destory sources
+	if (auto* ss = session.try_get<Components::VoIP::StreamSources>(); ss != nullptr) {
+		for (const auto ssov : ss->streams) {
+
+			_os.throwEventDestroy(ssov);
+			_os.registry().destroy(ssov);
+		}
+	}
+
+	// destory sinks
+	if (auto* ss = session.try_get<Components::VoIP::StreamSinks>(); ss != nullptr) {
+		for (const auto ssov : ss->streams) {
+
+			_os.throwEventDestroy(ssov);
+			_os.registry().destroy(ssov);
+		}
+	}
+
+	// destory session
+	_os.throwEventDestroy(session);
+	_os.registry().destroy(session);
+}
+
+ToxAVVoIPModel::ToxAVVoIPModel(ObjectStore2& os, ToxAVI& av, Contact3Registry& cr, ToxContactModel2& tcm) :
+	_os(os), _av(av), _cr(cr), _tcm(tcm)
+{
+	_av.subscribe(this, ToxAV_Event::friend_call);
+	_av.subscribe(this, ToxAV_Event::friend_call_state);
+	_av.subscribe(this, ToxAV_Event::friend_audio_bitrate);
+	_av.subscribe(this, ToxAV_Event::friend_video_bitrate);
+	_av.subscribe(this, ToxAV_Event::friend_audio_frame);
+	_av.subscribe(this, ToxAV_Event::friend_video_frame);
+
+	// attach to all tox friend contacts
+
+	for (const auto& [cv, _] : _cr.view<Contact::Components::ToxFriendPersistent>().each()) {
+		_cr.emplace<VoIPModelI*>(cv, this);
+	}
+	// TODO: events
+}
+
+ToxAVVoIPModel::~ToxAVVoIPModel(void) {
+	for (const auto& [ov, voipmodel] : _os.registry().view<VoIPModelI*>().each()) {
+		if (voipmodel == this) {
+			destroySession(_os.objectHandle(ov));
+		}
+	}
+}
+
+void ToxAVVoIPModel::tick(void) {
+	for (const auto& [oc, asink] : _os.registry().view<ToxAVCallAudioSink*>().each()) {
+		if (!asink->_writer) {
+			continue;
+		}
+
+		for (size_t i = 0; i < 100; i++) {
+			auto new_frame_opt = asink->_writer->pop();
+			if (!new_frame_opt.has_value()) {
+				break;
+			}
+			const auto& new_frame = new_frame_opt.value();
+
+			//* @param sample_count Number of samples in this frame. Valid numbers here are
+			//*   `((sample rate) * (audio length) / 1000)`, where audio length can be
+			//*   2.5, 5, 10, 20, 40 or 60 milliseconds.
+
+			// we likely needs to subdivide/repackage
+			// frame size should be an option exposed to the user
+			// with 10ms as a default ?
+			// the larger the frame size, the less overhead but the more delay
+
+			auto err = _av.toxavAudioSendFrame(
+				asink->_fid,
+				new_frame.getSpan().ptr,
+				new_frame.getSpan().size / new_frame.channels,
+				new_frame.channels,
+				new_frame.sample_rate
+			);
+			if (err != TOXAV_ERR_SEND_FRAME_OK) {
+				std::cerr << "DTC: failed to send audio frame " << err << "\n";
+			}
+		}
+	}
+
+	for (const auto& [oc, vsink] : _os.registry().view<ToxAVCallVideoSink*>().each()) {
+		if (!vsink->_writer) {
+			continue;
+		}
+
+		for (size_t i = 0; i < 10; i++) {
+			auto new_frame_opt = vsink->_writer->pop();
+			if (!new_frame_opt.has_value()) {
+				break;
+			}
+			const auto& new_frame = new_frame_opt.value();
+
+			if (!new_frame.surface) {
+				// wtf?
+				continue;
+			}
+
+			// conversion is done in the sink's stream
+			SDL_Surface* surf = new_frame.surface.get();
+			assert(surf != nullptr);
+
+			SDL_LockSurface(surf);
+			_av.toxavVideoSendFrame(
+				vsink->_fid,
+				surf->w, surf->h,
+				static_cast<const uint8_t*>(surf->pixels),
+				static_cast<const uint8_t*>(surf->pixels) + surf->w * surf->h,
+				static_cast<const uint8_t*>(surf->pixels) + surf->w * surf->h + (surf->w/2) * (surf->h/2)
+			);
+			SDL_UnlockSurface(surf);
+		}
+	}
+}
+
+ObjectHandle ToxAVVoIPModel::enter(const Contact3 c, const Components::VoIP::DefaultConfig& defaults) {
+	if (!_cr.all_of<Contact::Components::ToxFriendEphemeral>(c)) {
+		return {};
+	}
+
+	const auto friend_number = _cr.get<Contact::Components::ToxFriendEphemeral>(c).friend_number;
+
+	const auto err = _av.toxavCall(friend_number, 0, 0);
+	if (err != TOXAV_ERR_CALL_OK) {
+		std::cerr << "TAVVOIP error: failed to start call: " << err << "\n";
+		return {};
+	}
+
+	ObjectHandle new_session {_os.registry(), _os.registry().create()};
+
+	new_session.emplace<VoIPModelI*>(this);
+	new_session.emplace<Components::VoIP::TagVoIPSession>(); // ??
+	new_session.emplace<Components::VoIP::SessionContact>(c);
+	new_session.emplace<Components::VoIP::SessionState>().state = Components::VoIP::SessionState::State::RINGING;
+	new_session.emplace<Components::VoIP::DefaultConfig>(defaults);
+
+	_os.throwEventConstruct(new_session);
+	return new_session;
+}
+
+bool ToxAVVoIPModel::accept(ObjectHandle session, const Components::VoIP::DefaultConfig& defaults) {
+	if (!static_cast<bool>(session)) {
+		return false;
+	}
+
+	if (!session.all_of<
+		Components::VoIP::TagVoIPSession,
+		VoIPModelI*,
+		Components::VoIP::SessionContact,
+		Components::VoIP::Incoming
+	>()) {
+		return false;
+	}
+
+	// check if self
+	if (session.get<VoIPModelI*>() != this) {
+		return false;
+	}
+
+	const auto session_contact = session.get<Components::VoIP::SessionContact>().c;
+	if (!_cr.all_of<Contact::Components::ToxFriendEphemeral>(session_contact)) {
+		return false;
+	}
+
+	const auto friend_number = _cr.get<Contact::Components::ToxFriendEphemeral>(session_contact).friend_number;
+	auto err = _av.toxavAnswer(friend_number, 0, 0);
+	if (err != TOXAV_ERR_ANSWER_OK) {
+		std::cerr << "TOXAVVOIP error: ansering call failed: " << err << "\n";
+		// we simply let it be for now, it apears we can try ansering later again
+		// we also get an error here when the call is already in progress (:
+		return false;
+	}
+
+	session.emplace<Components::VoIP::DefaultConfig>(defaults);
+
+	// answer defaults to enabled receiving audio and video
+	// TODO: think about how we should handle this
+	// set to disabled? and enable on src connection?
+	// we already default disabled send and enabled on sink connection
+	//_av.toxavCallControl(friend_number, TOXAV_CALL_CONTROL_HIDE_VIDEO);
+	//_av.toxavCallControl(friend_number, TOXAV_CALL_CONTROL_MUTE_AUDIO);
+
+
+	// how do we know the other side is accepting audio
+	// bitrate cb or what?
+	assert(!session.all_of<Components::ToxAVAudioSink>());
+	addAudioSink(session, friend_number);
+	assert(!session.all_of<Components::ToxAVVideoSink>());
+	addVideoSink(session, friend_number);
+
+	if (const auto* i_av = session.try_get<Components::ToxAVIncomingAV>(); i_av != nullptr) {
+		// create audio src
+		if (i_av->incoming_audio) {
+			assert(!session.all_of<Components::ToxAVAudioSource>());
+			addAudioSource(session, friend_number);
+		}
+
+		// create video src
+		if (i_av->incoming_video) {
+			assert(!session.all_of<Components::ToxAVVideoSource>());
+			addVideoSource(session, friend_number);
+		}
+	}
+
+	session.get_or_emplace<Components::VoIP::SessionState>().state = Components::VoIP::SessionState::State::CONNECTED;
+	_os.throwEventUpdate(session);
+	return true;
+}
+
+bool ToxAVVoIPModel::leave(ObjectHandle session) {
+	// rename to end?
+
+	if (!static_cast<bool>(session)) {
+		return false;
+	}
+
+	if (!session.all_of<
+		Components::VoIP::TagVoIPSession,
+		VoIPModelI*,
+		Components::VoIP::SessionContact
+	>()) {
+		return false;
+	}
+
+	// check if self
+	if (session.get<VoIPModelI*>() != this) {
+		return false;
+	}
+
+	const auto session_contact = session.get<Components::VoIP::SessionContact>().c;
+	if (!_cr.all_of<Contact::Components::ToxFriendEphemeral>(session_contact)) {
+		return false;
+	}
+
+	const auto friend_number = _cr.get<Contact::Components::ToxFriendEphemeral>(session_contact).friend_number;
+	// check error? (we delete anyway)
+	_av.toxavCallControl(friend_number, Toxav_Call_Control::TOXAV_CALL_CONTROL_CANCEL);
+
+	destroySession(session);
+	return true;
+}
+
+bool ToxAVVoIPModel::onEvent(const Events::FriendCall& e) {
+	// new incoming call, create voip session, ready to be accepted
+	// (or rejected...)
+
+	const auto session_contact = _tcm.getContactFriend(e.friend_number);
+	if (!_cr.valid(session_contact)) {
+		return false;
+	}
+
+	ObjectHandle new_session {_os.registry(), _os.registry().create()};
+
+	new_session.emplace<VoIPModelI*>(this);
+	new_session.emplace<Components::VoIP::TagVoIPSession>(); // ??
+	new_session.emplace<Components::VoIP::Incoming>(session_contact); // in 1on1 its always the same contact, might leave blank
+	new_session.emplace<Components::VoIP::SessionContact>(session_contact);
+	new_session.emplace<Components::VoIP::SessionState>().state = Components::VoIP::SessionState::State::RINGING;
+	new_session.emplace<Components::ToxAVIncomingAV>(e.audio_enabled, e.video_enabled);
+
+	_os.throwEventConstruct(new_session);
+	return true;
+}
+
+bool ToxAVVoIPModel::onEvent(const Events::FriendCallState& e) {
+	const auto session_contact = _tcm.getContactFriend(e.friend_number);
+	if (!_cr.valid(session_contact)) {
+		return false;
+	}
+
+	ToxAVFriendCallState s{e.state};
+
+	// find session(s?)
+	// TODO: keep lookup table
+	for (const auto& [ov, voipmodel] : _os.registry().view<VoIPModelI*>().each()) {
+		if (voipmodel == this) {
+			auto o = _os.objectHandle(ov);
+
+			if (!o.all_of<Components::VoIP::SessionContact>()) {
+				continue;
+			}
+			if (session_contact != o.get<Components::VoIP::SessionContact>().c) {
+				continue;
+			}
+
+			if (s.is_error() || s.is_finished()) {
+				// destroy call
+				destroySession(o);
+			} else {
+				// remote accepted our call, or av send/recv conditions changed?
+				o.get<Components::VoIP::SessionState>().state = Components::VoIP::SessionState::State::CONNECTED; // set to in call ??
+
+				if (s.is_accepting_a() && !o.all_of<Components::ToxAVAudioSink>()) {
+					addAudioSink(o, e.friend_number);
+				} else if (!s.is_accepting_a() && o.all_of<Components::ToxAVAudioSink>()) {
+					// remove asink?
+				}
+
+				// video
+				if (s.is_accepting_v() && !o.all_of<Components::ToxAVVideoSink>()) {
+					addVideoSink(o, e.friend_number);
+				} else if (!s.is_accepting_v() && o.all_of<Components::ToxAVVideoSink>()) {
+					// remove vsink?
+				}
+
+				// add/update sources
+				// audio
+				if (s.is_sending_a() && !o.all_of<Components::ToxAVAudioSource>()) {
+					addAudioSource(o, e.friend_number);
+				} else if (!s.is_sending_a() && o.all_of<Components::ToxAVAudioSource>()) {
+					// remove asrc?
+				}
+
+				// video
+				if (s.is_sending_v() && !o.all_of<Components::ToxAVVideoSource>()) {
+					addVideoSource(o, e.friend_number);
+				} else if (!s.is_sending_v() && o.all_of<Components::ToxAVVideoSource>()) {
+					// remove vsrc?
+				}
+			}
+		}
+	}
+
+	return true;
+}
+
+bool ToxAVVoIPModel::onEvent(const Events::FriendAudioBitrate&) {
+	return false;
+}
+
+bool ToxAVVoIPModel::onEvent(const Events::FriendVideoBitrate&) {
+	return false;
+}
+
+bool ToxAVVoIPModel::onEvent(const Events::FriendAudioFrame& e) {
+	auto asrc_it = _audio_sources.find(e.friend_number);
+	if (asrc_it == _audio_sources.cend()) {
+		// missing src from lookup table
+		return false;
+	}
+
+	auto asrc = asrc_it->second;
+
+	if (!static_cast<bool>(asrc)) {
+		// missing src to put frame into ??
+		return false;
+	}
+
+	assert(asrc.all_of<FrameStream2MultiSource<AudioFrame2>*>());
+	assert(asrc.all_of<Components::FrameStream2Source<AudioFrame2>>());
+
+	asrc.get<FrameStream2MultiSource<AudioFrame2>*>()->push(AudioFrame2{
+		e.sampling_rate,
+		e.channels,
+		std::vector<int16_t>(e.pcm.begin(), e.pcm.end()) // copy
+	});
+
+	return true;
+}
+
+bool ToxAVVoIPModel::onEvent(const Events::FriendVideoFrame& e) {
+	auto vsrc_it = _video_sources.find(e.friend_number);
+	if (vsrc_it == _video_sources.cend()) {
+		// missing src from lookup table
+		return false;
+	}
+
+	auto vsrc = vsrc_it->second;
+
+	if (!static_cast<bool>(vsrc)) {
+		// missing src to put frame into ??
+		return false;
+	}
+
+	assert(vsrc.all_of<FrameStream2MultiSource<SDLVideoFrame>*>());
+	assert(vsrc.all_of<Components::FrameStream2Source<SDLVideoFrame>>());
+
+	auto* new_surf = SDL_CreateSurface(e.width, e.height, SDL_PIXELFORMAT_IYUV);
+	assert(new_surf);
+	if (SDL_LockSurface(new_surf)) {
+		// copy the data
+		// we know how the implementation works, its y u v consecutivlely
+		// y
+		for (size_t y = 0; y < e.height; y++) {
+			std::memcpy(
+				//static_cast<uint8_t*>(new_surf->pixels) + new_surf->pitch*y,
+				static_cast<uint8_t*>(new_surf->pixels) + e.width*y,
+				e.y.ptr + e.ystride*y,
+				e.width
+			);
+		}
+
+		// u
+		for (size_t y = 0; y < e.height/2; y++) {
+			std::memcpy(
+				static_cast<uint8_t*>(new_surf->pixels) + (e.width*e.height) + (e.width/2)*y,
+				e.u.ptr + e.ustride*y,
+				e.width/2
+			);
+		}
+
+		// v
+		for (size_t y = 0; y < e.height/2; y++) {
+			std::memcpy(
+				static_cast<uint8_t*>(new_surf->pixels) + (e.width*e.height) + ((e.width/2)*(e.height/2)) + (e.width/2)*y,
+				e.v.ptr + e.vstride*y,
+				e.width/2
+			);
+		}
+
+		SDL_UnlockSurface(new_surf);
+	}
+
+	vsrc.get<FrameStream2MultiSource<SDLVideoFrame>*>()->push({
+		// ms -> us
+		Message::getTimeMS() * 1000, // TODO: make more precise
+		new_surf
+	});
+
+	SDL_DestroySurface(new_surf);
+	return false;
+}
+

src/tox_av_voip_model.hpp

@@ -0,0 +1,48 @@
+#pragma once
+
+#include <solanaceae/object_store/fwd.hpp>
+#include <solanaceae/contact/contact_model3.hpp>
+#include <solanaceae/tox_contacts/tox_contact_model2.hpp>
+#include "./frame_streams/voip_model.hpp"
+#include "./tox_av.hpp"
+
+#include <unordered_map>
+
+class ToxAVVoIPModel : protected ToxAVEventI, public VoIPModelI {
+	ObjectStore2& _os;
+	ToxAVI& _av;
+	Contact3Registry& _cr;
+	ToxContactModel2& _tcm;
+
+	// for faster lookup
+	std::unordered_map<uint32_t, ObjectHandle> _audio_sources;
+	std::unordered_map<uint32_t, ObjectHandle> _video_sources;
+
+	// TODO: virtual? strategy? protected?
+	virtual void addAudioSource(ObjectHandle session, uint32_t friend_number);
+	virtual void addAudioSink(ObjectHandle session, uint32_t friend_number);
+	virtual void addVideoSource(ObjectHandle session, uint32_t friend_number);
+	virtual void addVideoSink(ObjectHandle session, uint32_t friend_number);
+
+	void destroySession(ObjectHandle session);
+
+	public:
+		ToxAVVoIPModel(ObjectStore2& os, ToxAVI& av, Contact3Registry& cr, ToxContactModel2& tcm);
+		~ToxAVVoIPModel(void);
+
+		void tick(void);
+
+	public: // voip model
+		ObjectHandle enter(const Contact3 c, const Components::VoIP::DefaultConfig& defaults) override;
+		bool accept(ObjectHandle session, const Components::VoIP::DefaultConfig& defaults) override;
+		bool leave(ObjectHandle session) override;
+
+	protected: // toxav events
+		bool onEvent(const Events::FriendCall&) override;
+		bool onEvent(const Events::FriendCallState&) override;
+		bool onEvent(const Events::FriendAudioBitrate&) override;
+		bool onEvent(const Events::FriendVideoBitrate&) override;
+		bool onEvent(const Events::FriendAudioFrame&) override;
+		bool onEvent(const Events::FriendVideoFrame&) override;
+};
+