diff --git a/src/main_screen.cpp b/src/main_screen.cpp
index 0b9c3b7..380d8e2 100644
--- a/src/main_screen.cpp
+++ b/src/main_screen.cpp
@@ -647,7 +647,7 @@ Screen* MainScreen::tick(float time_delta, bool& quit) {
 	//const float av_interval = std::pow(tav.toxavIterationInterval(), 1.18)/1000.f;
 	const float av_interval = tav.toxavIterationInterval()/1000.f;
 
-	tavvoip.tick();
+	const float av_voip_interval = tavvoip.tick();
 #endif
 
 	tcm.iterate(time_delta); // compute
@@ -694,6 +694,10 @@ Screen* MainScreen::tick(float time_delta, bool& quit) {
 		_min_tick_interval,
 		av_interval
 	);
+	_min_tick_interval = std::min<float>(
+		_min_tick_interval,
+		av_voip_interval
+	);
 #endif
 
 	//std::cout << "MS: min tick interval: " << _min_tick_interval << "\n";
diff --git a/src/tox_av.hpp b/src/tox_av.hpp
index 34746a9..0791513 100644
--- a/src/tox_av.hpp
+++ b/src/tox_av.hpp
@@ -109,6 +109,9 @@ struct ToxAVI : public ToxAVEventProviderI {
 	ToxAVI(Tox* tox);
 	virtual ~ToxAVI(void);
 
+	// NOTE: interval timers are only interesting for receiving streams
+	// if we are only sending, it will always report 195ms
+
 	// interface
 	// if iterate is called on a different thread, it will fire events there
 	uint32_t toxavIterationInterval(void) const;
diff --git a/src/tox_av_voip_model.cpp b/src/tox_av_voip_model.cpp
index 7ad2f40..4951b0c 100644
--- a/src/tox_av_voip_model.cpp
+++ b/src/tox_av_voip_model.cpp
@@ -104,6 +104,7 @@ struct ToxAVCallVideoSink : public FrameStream2SinkI<SDLVideoFrame> {
 
 	uint32_t _fid;
 	std::shared_ptr<stream_type> _writer;
+	uint64_t _last_ts {0};
 
 	ToxAVCallVideoSink(ToxAVI& toxav, uint32_t fid) : _toxav(toxav), _fid(fid) {}
 	~ToxAVCallVideoSink(void) {
@@ -139,6 +140,7 @@ struct ToxAVCallVideoSink : public FrameStream2SinkI<SDLVideoFrame> {
 
 		if (sub == _writer) {
 			_writer = nullptr;
+			_last_ts = 0;
 
 			/*auto err = */_toxav.toxavVideoSetBitRate(_fid, 0);
 			// print warning? on error?
@@ -350,7 +352,7 @@ void ToxAVVoIPModel::audio_thread_tick(void) {
 			//*   `((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
+			// we likely need 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
@@ -372,7 +374,7 @@ void ToxAVVoIPModel::audio_thread_tick(void) {
 void ToxAVVoIPModel::video_thread_tick(void) {
 	//for (const auto& [oc, vsink] : _os.registry().view<ToxAVCallVideoSink*>().each()) {
 	std::lock_guard lg{_video_sinks_mutex};
-	for (const auto* vsink : _video_sinks) {
+	for (auto* vsink : _video_sinks) {
 		if (!vsink->_writer) {
 			continue;
 		}
@@ -389,6 +391,15 @@ void ToxAVVoIPModel::video_thread_tick(void) {
 				continue;
 			}
 
+			// interval estimate based on 2 frames
+			if (vsink->_last_ts != 0 && vsink->_last_ts < new_frame.timestampUS) {
+				const auto interval_us = new_frame.timestampUS - vsink->_last_ts;
+				if (_video_recent_interval > interval_us) {
+					_video_recent_interval = interval_us;
+				}
+			}
+			vsink->_last_ts = new_frame.timestampUS;
+
 			// conversion is done in the sink's stream
 			SDL_Surface* surf = new_frame.surface.get();
 			assert(surf != nullptr);
@@ -519,7 +530,7 @@ ToxAVVoIPModel::~ToxAVVoIPModel(void) {
 	}
 }
 
-void ToxAVVoIPModel::tick(void) {
+float ToxAVVoIPModel::tick(void) {
 	std::lock_guard lg{_e_queue_mutex};
 	while (!_e_queue.empty()) {
 		const auto& e_var = _e_queue.front();
@@ -536,6 +547,14 @@ void ToxAVVoIPModel::tick(void) {
 
 		_e_queue.pop_front();
 	}
+
+	// TODO: audio (ez, just do 60ms if sending audio)
+	auto interval = _video_recent_interval/1'000'000.f;
+
+	// funky if tickrate momentarily higher than frame rate
+	_video_recent_interval = 10'000'000;
+
+	return interval;
 }
 
 ObjectHandle ToxAVVoIPModel::enter(const Contact4 c, const Components::VoIP::DefaultConfig& defaults) {
diff --git a/src/tox_av_voip_model.hpp b/src/tox_av_voip_model.hpp
index 5a1ceae..6aa9f6e 100644
--- a/src/tox_av_voip_model.hpp
+++ b/src/tox_av_voip_model.hpp
@@ -10,6 +10,7 @@
 #include <variant>
 #include <deque>
 #include <mutex>
+#include <atomic>
 
 // fwd
 struct ToxAVCallAudioSink;
@@ -24,7 +25,7 @@ class ToxAVVoIPModel : protected ToxAVEventI, public VoIPModelI {
 
 	uint64_t _pad0;
 	// these events need to be worked on the main thread instead
-	// TODO: replac ewith lockless queue
+	// TODO: replace with lockless queue
 	std::deque<
 	std::variant<
 		Events::FriendCall,
@@ -42,6 +43,11 @@ class ToxAVVoIPModel : protected ToxAVEventI, public VoIPModelI {
 	std::mutex _video_sinks_mutex;
 	uint64_t _pad3;
 
+	// filled with min() in video_thread_tick()
+	// ms, 10sec means none
+	std::atomic<uint64_t> _video_recent_interval{10'000'000};
+	uint64_t _pad4;
+
 	// for faster lookup
 	std::unordered_map<uint32_t, ObjectHandle> _audio_sources;
 	std::unordered_map<uint32_t, ObjectHandle> _video_sources;
@@ -54,8 +60,7 @@ class ToxAVVoIPModel : protected ToxAVEventI, public VoIPModelI {
 
 	void destroySession(ObjectHandle session);
 
-	// TODO: this needs to move to the toxav thread
-	// we could use "events" as pre/post audio/video iterate...
+	// we use "events" as pre/post audio/video iterate...
 	void audio_thread_tick(void);
 	void video_thread_tick(void);
 
@@ -67,7 +72,7 @@ class ToxAVVoIPModel : protected ToxAVEventI, public VoIPModelI {
 		~ToxAVVoIPModel(void);
 
 		// handle events coming from toxav thread(s)
-		void tick(void);
+		float tick(void);
 
 	public: // voip model
 		ObjectHandle enter(const Contact4 c, const Components::VoIP::DefaultConfig& defaults) override;