Preserve stream capture settings and refine audio playback

src/stream/stream_playback_audio.cpp

@@ -30,6 +30,7 @@ struct stream_playback_audio {
     soundtouch::SoundTouch stretch;
     Span input[mapCapacity], output[mapCapacity];
     unsigned inputHead = 0, inputCount = 0, outputHead = 0, outputCount = 0;
+    unsigned bufferFrames = 0;
     int tempo = 1000;
     int64_t startPts = 0, expectedPts = -1;
     double expectedOutput = 0, producedOutput = 0, lastPts = 0;
@@ -85,12 +86,15 @@ static int feedPcm(stream_playback_audio* a, int64_t pts, const uint8_t* packet,
 
 extern "C" {
 
-struct stream_playback_audio* stream_playback_audio_create(int64_t start) {
+struct stream_playback_audio* stream_playback_audio_create(int64_t start, unsigned bufferMs) {
     if (start < 0) { DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "playback audio invalid start=%lld", (long long)start); return nullptr; }
+    if (!bufferMs || bufferMs > 1000) {
+        DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "playback audio invalid PCM buffer=%ums range=1..1000ms", bufferMs); return nullptr;
+    }
     void* memory = u_malloc(sizeof(stream_playback_audio));
     if (!memory) { DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "playback audio allocation failed"); return nullptr; }
     auto* a = new (memory) stream_playback_audio;
-    a->startPts = start; a->lastPts = (double)start;
+    a->startPts = start; a->lastPts = (double)start; a->bufferFrames = rate * bufferMs / 1000;
     a->opus = opus_codec_decoder_create(rate, 2);
     if (!a->opus) {
         DEBUG_ERROR(DEBUG_CATEGORY_STREAM, "playback audio Opus initialization failed"); stream_playback_audio_destroy(a); return nullptr;
@@ -99,7 +103,8 @@ struct stream_playback_audio* stream_playback_audio_create(int64_t start) {
     a->stretch.setSetting(SETTING_USE_QUICKSEEK, 1);
     a->stretch.setSetting(SETTING_SEQUENCE_MS, 20); a->stretch.setSetting(SETTING_SEEKWINDOW_MS, 10);
     a->stretch.setSetting(SETTING_OVERLAP_MS, 4); a->stretch.setSetting(SETTING_USE_AA_FILTER, 0);
-    DEBUG_INFO(DEBUG_CATEGORY_STREAM, "playback audio created start=%lld stereo=1 tempo=1.0..1.1", (long long)start);
+    DEBUG_INFO(DEBUG_CATEGORY_STREAM, "playback audio created start=%lld stereo=1 tempo=1.0..1.1 PCM_buffer=%ums/%u_frames",
+               (long long)start, bufferMs, a->bufferFrames);
     return a;
 }
 
@@ -143,7 +148,7 @@ void stream_playback_audio_set_tempo(struct stream_playback_audio* a, int tempo)
 }
 
 int stream_playback_audio_needs_input(const struct stream_playback_audio* a) {
-    return a && !a->finished && !a->failed && a->stretch.numSamples() < framesPerPacket;
+    return a && !a->finished && !a->failed && a->stretch.numSamples() < a->bufferFrames;
 }
 
 /* Ограничение готовой границы применяется до выдачи PCM устройству, сохраняя общий A/V clock. */

src/stream/stream_playback_audio.h

@@ -15,7 +15,9 @@ extern "C" {
 #endif
 
 struct stream_playback_audio;
-struct stream_playback_audio *stream_playback_audio_create(int64_t start_pts_us);
+/* buffer_ms (1..1000) — порог пополнения готового PCM; трансляции используют 40 мс.
+ * Это запас декодера, без аппаратной задержки и сетевого джиттер-буфера. */
+struct stream_playback_audio *stream_playback_audio_create(int64_t start_pts_us, unsigned buffer_ms);
 void stream_playback_audio_destroy(struct stream_playback_audio *audio);
 int stream_playback_audio_feed(struct stream_playback_audio *audio, int64_t pts_us, const uint8_t *packet, size_t length);
 void stream_playback_audio_set_tempo(struct stream_playback_audio *audio, int tempo_milli);