audio #388

app/wyzebridge/ffmpeg.py

@@ -29,19 +29,19 @@ def get_ffmpeg_cmd(
     - list of str: complete ffmpeg command that is ready to run as subprocess.
     """
 
-    flags = "-fflags +flush_packets+nobuffer -flags +low_delay+global_header"
+    flags = "-fflags +flush_packets+nobuffer+genpts -flags +low_delay+global_header"
     livestream = get_livestream_cmd(uri)
     audio_in = "-f lavfi -i anullsrc=cl=mono" if livestream else ""
     audio_out = "aac"
-    thread_queue = "-thread_queue_size 1k -analyzeduration 32 -probesize 32"
+    thread_queue = "-thread_queue_size 64 -analyzeduration 32 -probesize 32"
     if audio and "codec" in audio:
-        audio_in += f"{thread_queue} -f {audio['codec']} -ac 1 -ar {audio['rate']} -i /tmp/{uri}_audio.pipe"
+        audio_in = f"{thread_queue} -f {audio['codec']} -ac 1 -ar {audio['rate']} -i /tmp/{uri}_audio.pipe"
         audio_out = audio["codec_out"] or "copy"
-    a_filter = env_bool("AUDIO_FILTER", "volume=5")
+    a_filter = env_bool("AUDIO_FILTER", "volume=5") + ",adelay=0|0"
     a_options = ["-compression_level", "4", "-filter:a", a_filter]
     rtsp_transport = "udp" if "udp" in env_bool("MTX_PROTOCOLS") else "tcp"
-    fio_cmd = "use_fifo=1:fifo_options=queue_size=250\\\:attempt_recovery=1\\\:drop_pkts_on_overflow=1"
-    rss_cmd = f"[{fio_cmd}:{{}}f=rtsp:{rtsp_transport=:}:bsfs/v=dump_extra=freq=k]rtsp://0.0.0.0:8554/{uri}"
+    fio_cmd = "use_fifo=1:fifo_options=attempt_recovery=1\\\:drop_pkts_on_overflow=1:"
+    rss_cmd = f"[{fio_cmd}{{}}f=rtsp:{rtsp_transport=:}:bsfs=dump_extra=freq=e]rtsp://0.0.0.0:8554/{uri}"
     rtsp_ss = rss_cmd.format("")
     if env_cam("AUDIO_STREAM", uri, style="original") and audio:
         rtsp_ss += "|" + rss_cmd.format("select=a:") + "_audio"
@@ -56,15 +56,12 @@ def get_ffmpeg_cmd(
         + (["-hwaccel", h264_enc] if h264_enc in {"vaapi", "qsv"} else [])
         + ["-f", vcodec, "-i", "pipe:0"]
         + audio_in.split()
-        + ["-c:v"]
+        + ["-map", "0:v", "-c:v"]
         + re_encode_video(uri, is_vertical)
-        + (["-c:a", audio_out] if audio_in else [])
+        + (["-map", "1:a", "-c:a", audio_out] if audio_in else [])
         + (a_options if audio and audio_out != "copy" else [])
-        + ["-fps_mode", "passthrough", "-flush_packets", "1"]
-        + ["-copyts", "-copytb", "0", "-async", "1"]
-        + ["-bsf:a", "setts='TS-STARTPTS'"]
-        + ["-map", "0:v"]
-        + (["-map", "1:a"] if audio_in else [])
+        + ["-fps_mode", "drop", "-async", "0", "-flush_packets", "1"]
+        + ["-muxdelay", "0", "-copytb", "1"]
         + ["-f", "tee"]
         + [rtsp_ss + get_record_cmd(uri, audio_out, record) + livestream]
     )

app/wyzecam/iotc.py

@@ -489,6 +489,7 @@ def recv_bridge_frame(self, timeout: int = 15, fps: int = 15) -> Iterator[bytes]
                     warnings.warn("Still waiting for first frame. Updating frame size.")
                     last["key_time"] = last["time"] = time.time()
                     self.update_frame_size_rate()
+                    yield b""
                     continue
                 self.state = WyzeIOTCSessionState.CONNECTING_FAILED
                 raise Exception(f"Stream did not receive a frame for over {timeout}s")
@@ -500,26 +501,31 @@ def recv_bridge_frame(self, timeout: int = 15, fps: int = 15) -> Iterator[bytes]
             if errno < 0:
                 time.sleep(sleep_interval)
                 if errno == tutk.AV_ER_DATA_NOREADY:
+                    yield b""
                     continue
                 if errno in (
                     tutk.AV_ER_INCOMPLETE_FRAME,
                     tutk.AV_ER_LOSED_THIS_FRAME,
                 ):
                     warnings.warn(str(tutk.TutkError(errno).name))
+                    yield b""
                     continue
                 raise tutk.TutkError(errno)
             if not frame_data:
+                yield b""
                 continue
             assert frame_info is not None, "Got no frame info without an error!"
             if frame_info.frame_size not in ignore_res:
                 if last["key_frame"] == 0:
                     warnings.warn(
                         f"Skipping smaller frame at start of stream (frame_size={frame_info.frame_size})"
                     )
+                    yield b""
                     continue
                 warnings.warn(f"Wrong resolution (frame_size={frame_info.frame_size})")
                 self.update_frame_size_rate()
                 last |= {"key_frame": 0, "key_time": 0, "time": time.time()}
+                yield b""
                 continue
             if frame_info.is_keyframe:
                 last |= {"key_frame": frame_info.frame_no, "key_time": time.time()}