Experiencing skipping when using output.file from an m3u8 of HLS chunks

[TheFrankman]

[SOLVED]

In a nutshell, i should have been using input.hls. I thought there was an issue with this, and the alpine docker image but it transpired it was just a typo in my M3u8 file.

Original Question

I am trying to concatenate hls fragments back into a single file, but the file that is produced has skips at the joins. Only slightly but they are audible.

My implementation is a big more complicated than this, as I have docker and docker-in-docker (dind) dynamically generating this stuff. But in essence I have one liquidsoap file that is continuously writing chunks to a centralised file storage, and a liquidsoap script that is responsible for concatenating a subsection of those chunks using an M3U8 file that i generate.

My hls liq looks something like this :

#!/usr/bin/liquidsoap

settings.init.allow_root.set(true)

PRIMARY_STREAM=getenv("PRIMARY_STREAM")
SECONDARY_STREAM=getenv("SECONDARY_STREAM")
PROVIDER=getenv("PROVIDER")
PRIMARY_STREAM_NAME=getenv("PRIMARY_STREAM_NAME")

log.stdout.set(true)

segment_duration=float_of_string(getenv(default="60.", "SEGMENT_DURATION"))

to_record = input.http(PRIMARY_STREAM)
silence = blank(duration=5.)

def segment_name(~position,~extname,stream_name) =
  timestamp = int_of_float(time())
  duration = int_of_float(segment_duration)
  "#{timestamp}.#{extname}"
end

if SECONDARY_STREAM != '' then
 secondary = input.http(SECONDARY_STREAM)
 to_record = fallback(track_sensitive=false, [to_record, secondary, silence])
else
 to_record = fallback(track_sensitive=false, [to_record, silence])
end

output.file.hls(fallible=true,
"/var/www/chunks/#{PROVIDER}/#{PRIMARY_STREAM_NAME}",
[("mp3-low", %mp3(bitrate=192))],
to_record,
segment_duration=60.,
segment_name=segment_name,
segments=4500,
persist_at="/var/www/chunks/#{PROVIDER}/#{PRIMARY_STREAM_NAME}.config",
)

Here is an example M3U8 file that i generate :

#EXT3U
#EXT-X-TARGETDURATION:60
#EXT-X-VERSION:3
#EXT-X-DISCONTINUITY-SEQUENCE:0
#EXT-X-MEDIA-SEQUENCE:0
#EXT-X-STREAM-INF:BANDWIDTH=192000,CODECS="mp4a.40.34"
#EXTINF:0.640,
/var/www/chunks/example/example.mp3/1678278038.mp3
#EXTINF:0.640,
/var/www/chunks/example/example.mp3/1678278098.mp3
#EXTINF:0.640,
/var/www/chunks/example/example.mp3/1678278158.mp3
#EXTINF:0.640,
/var/www/chunks/example/example.mp3/1678278218.mp3
#EXTINF:0.640,
/var/www/chunks/example/example.mp31678278278.mp3
#EXTINF:0.640,
/var/www/chunks/example/example.mp3/1678278338.mp3
#EXTINF:0.640,
/var/www/chunks/example/example.mp3/1678278398.mp3
#EXT-X-ENDLIST

and here is my liquidsoap script responsible for concat

#!/usr/bin/liquidsoap
settings.init.allow_root.set(true)
M3U_PATH=getenv("M3U_PATH")
OUTPUT_PATH=getenv("OUTPUT_PATH")

log.important("M3U PATH IS  #{M3U_PATH}")
log.important("OUTPUT PATH IS  #{OUTPUT_PATH}")

playlist_to_concat = playlist(M3U_PATH, loop=false, mode='normal')

clock.assign_new(sync="none",[playlist_to_concat])

output.file(
    %mp3(bitrate=192),
    OUTPUT_PATH,
    playlist_to_concat,
    fallible=true,
    on_stop=shutdown,
    duration=playlist_to_concat.duration()
)

My approach for the concat is that i load the playlist from the m3u8, i play it once, i dsync the clock and output to a file.

Can anyone help me identify where the issue may be here, i feel as though this should work but the skips are there.

Here is concatenate logs :

2023/03/08 16:06:14 >>> LOG START
2023/03/08 16:06:13 [ffmpeg.filter.bitstream:3] No valid mode found for filter pgs_frame_merge!
2023/03/08 16:06:13 [main:3] Liquidsoap 2.1.3
2023/03/08 16:06:13 [main:3] Using: bytes=[distributed with OCaml 4.02 or above] posix-time2=2.0.0 pcre=7.5.0 sedlex=3.0 menhirLib=20220210 curl=0.9.2 uri=4.2.0 memtrace=v0.2.2 mem_usage=0.0.4 dtools=0.4.4 duppy=0.9.3 cry=0.6.8 mm=0.8.2 xmlplaylist=0.1.5 lastfm=0.3.3 ogg=0.7.3 ogg.decoder=0.7.3 vorbis=0.8.1 vorbis.decoder=0.8.1 opus=0.2.3 opus.decoder=0.2.3 speex=0.4.1 speex.decoder=0.4.1 mad=0.5.2 flac=0.3.1 flac.ogg=0.3.1 flac.decoder=0.3.1 dynlink=[distributed with Ocaml] lame=0.3.6 shine=0.2.3 frei0r=0.1.2 fdkaac=0.3.2 theora=0.4.0 theora.decoder=0.4.0 ffmpeg=1.1.6 bjack=0.1.6 alsa=0.3.0 ao=0.2.4 samplerate=0.1.6 taglib=0.3.10 ssl=0.5.12 magic=0.7.3 camomile=1.0.2 inotify=2.0-58-g1beb568 faad=0.5.2 soundtouch=0.1.9 portaudio=0.2.3 pulseaudio=0.1.5 ladspa=0.2.2 tsdl=v0.9.9 tsdl-ttf=0.3.2 tsdl-image=0.3.2 camlimages=4.2.6 camlimages.freetype=5.0.4 cohttp-lwt-unix=5.0.0 prometheus-app=1.2 srt.constants=0.2.2 srt.types=0.2.2 srt.stubs=0.2.2 srt.stubs.locked=0.2.2 srt=0.2.2 lo=0.2.0 gd=1.0a5 irc-client-unix=[unspecified]
2023/03/08 16:06:13 [clock:3] Using native (high-precision) implementation for latency control
2023/03/08 16:06:14 [lang:3] M3U PATH IS  /var/www/recordings/example/example.mp3/1678290755-1678291055.m3u8
2023/03/08 16:06:14 [lang:3] OUTPUT PATH IS  /var/www/recordings/example/example.mp3/1678290755-1678291055.mp3
2023/03/08 16:06:14 [frame:3] Using 44100Hz audio, 25Hz video, 44100Hz main.
2023/03/08 16:06:14 [frame:3] Video frame size set to: 1280x720
2023/03/08 16:06:14 [frame:3] Frame size must be a multiple of 1764 ticks = 1764 audio samples = 1 video samples.
2023/03/08 16:06:14 [frame:3] Targeting 'frame.duration': 0.04s = 1764 audio samples = 1764 ticks.
2023/03/08 16:06:14 [frame:3] Frames last 0.04s = 1764 audio samples = 1 video samples = 1764 ticks.
2023/03/08 16:06:14 [sandbox:3] Sandboxing disabled
2023/03/08 16:06:14 [video.converter:3] Using preferred video converter: ffmpeg.
2023/03/08 16:06:14 [audio.converter:3] Using samplerate converter: libsamplerate.
2023/03/08 16:06:14 [video.text:3] Using camlimages implementation
2023/03/08 16:06:14 [clock.1678290755-1678291055.m3u8:3] Streaming loop starts in no sync mode
2023/03/08 16:06:14 [clock.1678290755-1678291055.m3u8:3] Delegating synchronisation to active sources
2023/03/08 16:06:15 [1678290755-1678291055.m3u8:3] Prepared "/var/www/chunks/example/example.mp3/1678290755.mp3" (RID 1).
2023/03/08 16:06:16 [decoder:2] Decoding "/var/www/chunks/example/example.mp3/1678290755.mp3" ended: Ffmpeg_decoder.End_of_file.
2023/03/08 16:06:16 [1678290755-1678291055.m3u8:3] Prepared "/var/www/chunks/example/example.mp3/1678290815.mp3" (RID 2).
2023/03/08 16:06:18 [decoder:2] Decoding "/var/www/chunks/example/example.mp3/1678290815.mp3" ended: Ffmpeg_decoder.End_of_file.
2023/03/08 16:06:18 [1678290755-1678291055.m3u8:3] Prepared "/var/www/chunks/example/example.mp3/1678290875.mp3" (RID 3).
[mp3 @ 0xffffb3977340] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977350] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977330] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977340] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3d99510] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3d99520] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977350] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977320] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffaf0004f0] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffae9293e0] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977330] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977340] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb098a9c0] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb098a9d0] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977350] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977320] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffae929680] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffae929690] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977330] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977340] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb0cce980] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb09ab070] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977350] Estimating duration from bitrate, this may be inaccurate
[mp3 @ 0xffffb3977320] Estimating duration from bitrate, this may be inaccurate
2023/03/08 16:06:19 [decoder:2] Decoding "/var/www/chunks/example/example.mp3/1678290875.mp3" ended: Ffmpeg_decoder.End_of_file.
2023/03/08 16:06:20 [1678290755-1678291055.m3u8:3] Prepared "/var/www/chunks/example/example.mp3/1678290935.mp3" (RID 4).
2023/03/08 16:06:21 [decoder:2] Decoding "/var/www/chunks/example/example.mp3/1678290935.mp3" ended: Ffmpeg_decoder.End_of_file.
2023/03/08 16:06:21 [1678290755-1678291055.m3u8:3] Prepared "/var/www/chunks/example/example.mp3/1678290995.mp3" (RID 5).
2023/03/08 16:06:24 [decoder:2] Decoding "/var/www/chunks/example/example.mp3/1678290995.mp3" ended: Ffmpeg_decoder.End_of_file.
2023/03/08 16:06:24 [1678290755-1678291055.m3u8:3] Prepared "/var/www/chunks/example/example.mp3/1678291055.mp3" (RID 6).
2023/03/08 16:06:25 [decoder:2] Decoding "/var/www/chunks/example/example.mp3/1678291055.mp3" ended: Ffmpeg_decoder.End_of_file.
2023/03/08 16:06:25 [/var/www/recordings/example/example.mp3/1678290755-1678291055.mp3:3] Source failed (no more tracks) stopping output...
2023/03/08 16:06:25 [main:3] Shutdown started!
2023/03/08 16:06:25 [main:3] Waiting for main threads to terminate...
2023/03/08 16:06:25 [clock.1678290755-1678291055.m3u8:3] Streaming loop stopped.
2023/03/08 16:06:25 [main:3] Main threads terminated.
2023/03/08 16:06:25 [threads:3] Shutting down scheduler...
2023/03/08 16:06:25 [threads:3] Scheduler shut down.
2023/03/08 16:06:25 [main:3] Cleaning downloaded files...
2023/03/08 16:06:25 [main:3] Freeing memory...

[TheFrankman]

I realise that this could be because of the estimated duration warning that's coming from FFMPEG, but i'm not sure how to work around that either.

[cdgraff]

I'm doing something similar, test with my settings.

  mp3 = 
    %ffmpeg(
      format="mp3",
      id3v2_version=0,
      write_xing=0,
      %audio(
        channels=2,
        samplerate=44100,
        codec="libmp3lame",
        b="64k"
      )
    )

  streams = [("mp3",mp3)]

[toots]

Hi! Have you tried using input.ffmpeg with the generated file?

[toots]

The reason is b/c the playlist will open/close on each segment which can indeed create gaps. input.ffmpeg should be able to detect the underlying HLS nature of the stream and decode the segments as part of a single logical mpegts/mp4 stream.

[TheFrankman]

FROM savonet/liquidsoap-alpine:v2.1.3

USER root
RUN apk update
RUN apk add bash-completion
RUN apk add neofetch
RUN apk add bubblewrap
RUN apk add vim

WORKDIR '/var/www/liquidsoap'

[toots]

That sounds about right. Only thing is, unless you run the docker image with special privileges, bubblewrap is useless. Also, the binary we ship does not need opam.

I'll try to have a try at reproducing your issue then.

[TheFrankman]

Yeah, i'm running with privileged: true and user: root. My docker compose file is built dynamically, but an example of what one looks like is :

version: "3.9"

services:
    station_name-1678726053-1678726173:
        build: .
        privileged: true
        user: root
        environment:
            M3U_PATH: "/var/www/recordings/customer/station_name/1678726053-1678726173.m3u8"
            OUTPUT_PATH: "/var/www/recordings/customer/station_name/1678726053-1678726173.mp3"
        volumes:
            - recorder-stack_shared-chunks:/var/www/chunks
            - recorder-stack_shared-recordings:/var/www/recordings
            - recorder-stack_shared-liquidsoap-src:/var/www/liquidsoap/src
        command: ["./src/concatenate.liq"]

volumes:
    recorder-stack_shared-chunks:
        external: true
    recorder-stack_shared-liquidsoap-src:
        external: true
    recorder-stack_shared-recordings:
        external: true

That being said, i can't remember why i added bubblewrap.

[TheFrankman]

Hello Toots,

I picked this project up again today and i spotted something really stupid. The error [input.ffmpeg_0:4] Connection failed: Avutil.Error(Invalid data found when processing input) was being caused due to a typo in my M3U8 file generation code.

The top line was #EXT3U when it should have been #EXTM3U bit embarrassing.

However i have established that input.hls is the way to go and it does work on the alpine docker image.

[toots]

Great glad this is sorted out.

[TheFrankman]

Hello @cdgraff ,

Thanks for your suggestion. I have added that format to both my hls generation liq and the conctenate liq but i am still seeing the skip in the output file :

I'll give the suggestion from @toots a go now

[cdgraff]

I tested and saw the same message but not into all the files

# ffmpeg -i "concat:1678296823.mp3|1678296853.mp3|1678296883.mp3" test.mp3
ffmpeg version 4.2.7-0ubuntu0.1 Copyright (c) 2000-2022 the FFmpeg developers
  built with gcc 9 (Ubuntu 9.4.0-1ubuntu1~20.04.1)
  configuration: --prefix=/usr --extra-version=0ubuntu0.1 --toolchain=hardened --libdir=/usr/lib/x86_64-linux-gnu --incdir=/usr/include/x86_64-linux-gnu --arch=amd64 --enable-gpl --disable-stripping --enable-avresample --disable-filter=resample --enable-avisynth --enable-gnutls --enable-ladspa --enable-libaom --enable-libass --enable-libbluray --enable-libbs2b --enable-libcaca --enable-libcdio --enable-libcodec2 --enable-libflite --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libgme --enable-libgsm --enable-libjack --enable-libmp3lame --enable-libmysofa --enable-libopenjpeg --enable-libopenmpt --enable-libopus --enable-libpulse --enable-librsvg --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libssh --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx265 --enable-libxml2 --enable-libxvid --enable-libzmq --enable-libzvbi --enable-lv2 --enable-omx --enable-openal --enable-opencl --enable-opengl --enable-sdl2 --enable-libdc1394 --enable-libdrm --enable-libiec61883 --enable-nvenc --enable-chromaprint --enable-frei0r --enable-libx264 --enable-shared
  libavutil      56. 31.100 / 56. 31.100
  libavcodec     58. 54.100 / 58. 54.100
  libavformat    58. 29.100 / 58. 29.100
  libavdevice    58.  8.100 / 58.  8.100
  libavfilter     7. 57.100 /  7. 57.100
  libavresample   4.  0.  0 /  4.  0.  0
  libswscale      5.  5.100 /  5.  5.100
  libswresample   3.  5.100 /  3.  5.100
  libpostproc    55.  5.100 / 55.  5.100
**[mp3 @ 0x564ddf3df280] Estimating duration from bitrate, this may be inaccurate**
Input #0, mp3, from 'concat:1678296823.mp3|1678296853.mp3|1678296883.mp3':
  Duration: 00:01:30.12, start: 0.000000, bitrate: 64 kb/s
    Stream #0:0: Audio: mp3, 44100 Hz, mono, fltp, 64 kb/s
Stream mapping:
  Stream #0:0 -> #0:0 (mp3 (mp3float) -> mp3 (libmp3lame))
Press [q] to stop, [?] for help
Output #0, mp3, to 'test.mp3':
  Metadata:
    TSSE            : Lavf58.29.100
    Stream #0:0: Audio: mp3 (libmp3lame), 44100 Hz, mono, fltp
    Metadata:
      encoder         : Lavc58.54.100 libmp3lame
size=     705kB time=00:01:30.12 bitrate=  64.0kbits/s speed=66.8x    
video:0kB audio:704kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.031476%

But listening the audio is perfect, no gaps...

[TheFrankman]

If i use ffmpeg directly, or even just cat the mp3 files together, it works without skips. It's only present when attempting to use output.file in LS. I think Toots is on the right path, it seems obvious that using playlist for my input would result in skips. So just need an alternative input.