diff --git a/libobs/media-io/audio-io.c b/libobs/media-io/audio-io.c
index dc93bfeaf36653..e923c86394061a 100644
--- a/libobs/media-io/audio-io.c
+++ b/libobs/media-io/audio-io.c
@@ -153,7 +153,7 @@ static inline void clamp_audio_output(struct audio_output *audio, size_t bytes)
 	}
 }
 
-static void input_and_output(struct audio_output *audio, uint64_t audio_time,
+static bool input_and_output(struct audio_output *audio, uint64_t audio_time,
 			     uint64_t prev_time)
 {
 	size_t bytes = AUDIO_OUTPUT_FRAMES * audio->block_size;
@@ -191,7 +191,7 @@ static void input_and_output(struct audio_output *audio, uint64_t audio_time,
 	success = audio->input_cb(audio->input_param, prev_time, audio_time,
 				  &new_ts, active_mixes, data);
 	if (!success)
-		return;
+		return false;
 
 	/* clamps audio data to -1.0..1.0 */
 	clamp_audio_output(audio, bytes);
@@ -199,6 +199,8 @@ static void input_and_output(struct audio_output *audio, uint64_t audio_time,
 	/* output */
 	for (size_t i = 0; i < MAX_AUDIO_MIXES; i++)
 		do_audio_output(audio, i, new_ts, AUDIO_OUTPUT_FRAMES);
+
+	return true;
 }
 
 static void *audio_thread(void *param)
@@ -217,6 +219,7 @@ static void *audio_thread(void *param)
 	uint32_t audio_wait_time =
 		(uint32_t)(audio_frames_to_ns(rate, AUDIO_OUTPUT_FRAMES) /
 			   1000000);
+	int wait_cycles = 0;
 
 	os_set_thread_name("audio-io: audio thread");
 
@@ -227,18 +230,31 @@ static void *audio_thread(void *param)
 	while (os_event_try(audio->stop_event) == EAGAIN) {
 		uint64_t cur_time;
 
-		os_sleep_ms(audio_wait_time);
+		if (wait_cycles > 0) {
+			os_sleep_ms(audio_wait_time / 2);
+			wait_cycles--;
+		} else {
+			os_sleep_ms(audio_wait_time);
+		}
 
 		profile_start(audio_thread_name);
 
 		cur_time = os_gettime_ns();
 		while (audio_time <= cur_time) {
-			samples += AUDIO_OUTPUT_FRAMES;
-			audio_time =
-				start_time + audio_frames_to_ns(rate, samples);
-
-			input_and_output(audio, audio_time, prev_time);
-			prev_time = audio_time;
+			uint64_t new_samples = samples + AUDIO_OUTPUT_FRAMES;
+			uint64_t new_audio_time =
+				start_time +
+				audio_frames_to_ns(rate, new_samples);
+
+			if (input_and_output(audio, new_audio_time,
+					     prev_time)) {
+				samples = new_samples;
+				audio_time = new_audio_time;
+				prev_time = audio_time;
+			} else {
+				os_sleep_ms(audio_wait_time / 2);
+				wait_cycles++;
+			}
 		}
 
 		profile_end(audio_thread_name);
diff --git a/libobs/obs-audio.c b/libobs/obs-audio.c
index e6b107b2b1286e..a787b51803e5a2 100644
--- a/libobs/obs-audio.c
+++ b/libobs/obs-audio.c
@@ -471,6 +471,7 @@ bool audio_callback(void *param, uint64_t start_ts_in, uint64_t end_ts_in,
 	struct ts_info ts = {start_ts_in, end_ts_in};
 	size_t audio_size;
 	uint64_t min_ts;
+	bool audio_pending = false;
 
 	da_resize(audio->render_order, 0);
 	da_resize(audio->root_nodes, 0);
@@ -516,6 +517,9 @@ bool audio_callback(void *param, uint64_t start_ts_in, uint64_t end_ts_in,
 		obs_source_audio_render(source, mixers, channels, sample_rate,
 					audio_size);
 
+		if (source->audio_pending && source->audio_ts)
+			audio_pending = true;
+
 		/* if a source has gone backward in time and we can no
 		 * longer buffer, drop some or all of its audio */
 		if (audio->total_buffering_ticks == MAX_BUFFERING_TICKS &&
@@ -611,5 +615,5 @@ bool audio_callback(void *param, uint64_t start_ts_in, uint64_t end_ts_in,
 	execute_audio_tasks();
 
 	UNUSED_PARAMETER(param);
-	return true;
+	return !audio_pending;
 }
diff --git a/libobs/obs-source.c b/libobs/obs-source.c
index d2c25cb1cd453c..3fc979ca65968c 100644
--- a/libobs/obs-source.c
+++ b/libobs/obs-source.c
@@ -1481,7 +1481,7 @@ static void source_output_audio_data(obs_source_t *source,
 
 	source->last_audio_ts = in.timestamp;
 	source->next_audio_ts_min =
-		in.timestamp + conv_frames_to_time(sample_rate, in.frames);
+		data->timestamp + conv_frames_to_time(sample_rate, in.frames);
 
 	in.timestamp += source->timing_adjust;