s302m in combination with loudnorm filter

[jb_alvarado]

Summary of the bug:
The audio codec s302m produce in combination with loudnorm filter an error:

[s302m @ 00000275a32cba00] number of samples in frame too big
Audio encoding failed

How to reproduce:

%  ffmpeg -f lavfi -i testsrc=size=1280x720:rate=25:duration=5 -f lavfi -i sine=frequency=1000:duration=5 -af "loudnorm=I=-23.0:TP=-1.5:LRA=11.0:dual_mono=true" -ac 2 -c:a s302m -strict -2 -f null -

Uncut output:

ffmpeg -f lavfi -i testsrc=size=1280x720:rate=25:duration=5 -f lavfi -i sine=frequency=1000:duration=5 -af "loudnorm=I=-23.0:TP=-1.5:LRA=11.0:dual_mono=true" -ac 2 -c:a s302m -strict -2 -f null -
ffmpeg version N-94713-g9c369b3222 Copyright (c) 2000-2019 the FFmpeg developers
  built with gcc 9.2.0 (Rev1, Built by MSYS2 project)
  configuration:  --enable-sdl2 --enable-gmp --enable-libmp3lame --enable-libvpx --enable-libx264 --enable-libx265 --enable-fontconfig --enable-libfreetype --enable-libmysof
a --enable-libopenjpeg --enable-libsoxr --enable-libtwolame --enable-libwavpack --enable-libwebp --enable-libxml2 --enable-libzimg --enable-gpl --enable-avisynth --enable-ch
romaprint --enable-libfdk-aac --enable-libfribidi --enable-librubberband --enable-libzmq --enable-opengl --enable-libsrt --enable-libaom --enable-schannel --extra-cflags=-DL
IBTWOLAME_STATIC --extra-cflags=-DCHROMAPRINT_NODLL --extra-libs=-lstdc++ --extra-cflags=-DZMQ_STATIC --extra-cflags=-DLIBXML_STATIC --enable-version3 --enable-nonfree --dis
able-stripping
  libavutil      56. 33.100 / 56. 33.100
  libavcodec     58. 56.100 / 58. 56.100
  libavformat    58. 31.104 / 58. 31.104
  libavdevice    58.  9.100 / 58.  9.100
  libavfilter     7. 58.101 /  7. 58.101
  libswscale      5.  6.100 /  5.  6.100
  libswresample   3.  6.100 /  3.  6.100
  libpostproc    55.  6.100 / 55.  6.100
Input #0, lavfi, from 'testsrc=size=1280x720:rate=25:duration=5':
  Duration: N/A, start: 0.000000, bitrate: N/A
    Stream #0:0: Video: rawvideo (RGB[24] / 0x18424752), rgb24, 1280x720 [SAR 1:1 DAR 16:9], 25 tbr, 25 tbn, 25 tbc
Input #1, lavfi, from 'sine=frequency=1000:duration=5':
  Duration: N/A, start: 0.000000, bitrate: 705 kb/s
    Stream #1:0: Audio: pcm_s16le, 44100 Hz, mono, s16, 705 kb/s
Stream mapping:
  Stream #0:0 -> #0:0 (rawvideo (native) -> wrapped_avframe (native))
  Stream #1:0 -> #0:1 (pcm_s16le (native) -> s302m (native))
Press [q] to stop, [?] for help
Output #0, null, to 'pipe:':
  Metadata:
    encoder         : Lavf58.31.104
    Stream #0:0: Video: wrapped_avframe, rgb24(progressive), 1280x720 [SAR 1:1 DAR 16:9], q=2-31, 200 kb/s, 25 fps, 25 tbn, 25 tbc
    Metadata:
      encoder         : Lavc58.56.100 wrapped_avframe
    Stream #0:1: Audio: s302m, 48000 Hz, stereo, s32 (24 bit), 2688 kb/s
    Metadata:
      encoder         : Lavc58.56.100 s302m
[Parsed_sine_0 @ 00000275a32aef00] EOF timestamp not reliable
[s302m @ 00000275a32cba00] number of samples in frame too big
Audio encoding failed
Conversion failed!

,asetnsamples=n=1024 could fix it, but produce randomly, depending from input, other errors and warnings.

[Elon Musk]

The limit in s302 is by design, max size of packet is 16bit number.
Can't you use -frame_size to set number of samples?

[jb_alvarado]

I tried -frame_size 1024, but it had no effect. When I use the filter asetnsamples=n=1024, it works with the command on top (sine source), but if I use a real audio source I get:

[out_0_0 @ 000002660d73d200] 100 buffers queued in out_0_0, something may be wrong.

In debug mode it looks also that not all packets are encoded:

cur_dts is invalid st:0 (0) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
    Last message repeated 139 times
[out_0_0 @ 0000021e3727e140] 100 buffers queued in out_0_0, something may be wrong.
[out_0_0 @ 0000021e3727e140] EOF on sink link out_0_0:default.
No more output streams to write to, finishing.
size=N/A time=00:00:25.12 bitrate=N/A speed=21.2x
video:0kB audio:8263kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown
Input file #0 (D:/stream/loudness/Sprache Leise.wav):
  Input stream #0:0 (audio): 1178 packets read (4823244 bytes); 1178 frames decoded (1205811 samples);
  Total: 1178 packets (4823244 bytes) demuxed
Output file #0 (pipe:):
  Output stream #0:0 (audio): 4712 frames encoded (1206016 samples); 4712 packets muxed (8460960 bytes);
  Total: 4712 packets (8460960 bytes) muxed
1178 frames successfully decoded, 0 decoding errors
[AVIOContext @ 0000021e3725e200] Statistics: 4888878 bytes read, 1 seeks

[Elon Musk]

I can not reproduce in any way issue with asetnsamples, please provide input file.

[jb_alvarado]

Ok, here the file. I use this command:

ffmpeg.exe -i "Sprache Leise.wav" -af "loudnorm=I=-23.0:TP=-1.5:LRA=11.0,asetnsamples=n=1024" -c:a s302m -strict -2 -f null -

[Elon Musk]

s302 encoder is marked as experimental for the reason that proper muxing need correct set for frame size in each packet.

100 queued buffers message is caused by bug/misfeature in loudnorm filter.

[jb_alvarado]

Yes I know is experimental, but s302 is the only supported lossless codec for mpegts. And in general works very nice in my project, better then mp2 and aac.

Is there a workaround to prevent this bug?

[jb_alvarado]

I think I have now the "workaround". The correct numbers of samples per frame needs to be calculated. For example for a 25 FPS Video: 192000 / 25 = 7680, this can be used in asetnsamples=7680, or in combination with aresample=48000: 48000 / 25 = 1920 -> asetnsamples=1920.

Is a bit confusing, because there is not always a video line and it also looks like that higher numbers like 8192 or also working...