[FFmpeg-user] Remuxing TS to MP4 causes black frame in Apple playback scenarios

[Sid Sethupathi]

Thanks Gyan - that seems to have worked.

I couldn't figure out how to do this via the command line but I updated my
program to reset the PTS to 0 on the input stream by reading the first
input packet's PTS and then subtracting that from all input packets' PTS.
Do you think that will have any unintended consequences, or is there maybe
a more idiomatic way to do this?

Doing the above seems to work on the few test cases I've tried: the output
mp4 has PTS = 0 on the first packet, the elst atom only has a single entry,
and the first frame does not show a black screen. I tried doing that same
"reset" logic after calling "av_interleaved_write_frame" but that didn't
work.

Sid



On Thu, May 2, 2024 at 12:47 AM Gyan Doshi <ffmpeg at gyani.pro> wrote:

>
>
> On 2024-05-02 03:21 am, Sid Sethupathi wrote:
> > Hello,
> >
> > I have a process that remuxes short HLS playlists into an mp4 format. The
> > output mp4 in this process has an all black first frame when played back
> in
> > either Safari or Quicktime Player. Everywhere else, the first frame does
> > not have this black frame.
> >
> > The process is an internal tool that uses libav directly so it's hard to
> > share the code, but I think I can reproduce via an ffmpeg command line
> > invocation against a single TS segment, so I'm hoping to be able to solve
> > the problem by understanding how to change the ffmpeg command.
> >
> > All the sample files can be found in this public Google Drive:
> >
> https://drive.google.com/drive/folders/1L8T5T4l_GUzxRnhjH-ojnT6-44FQLQq1?usp=drive_link
> > .
> >
> > bbb_2.ts is the second segment of Big Buck Bunny (sourced from this test
> > stream:
> >
> https://test-streams.mux.dev/x36xhzz/url_0/193039199_mp4_h264_aac_hd_7.m3u8
> ).
> > If you open this ts file in Quicktime player or Safari, the first frame
> is
> > NOT black.
> >
> > bbb_2_remuxed.mp4 has been remuxed using the following command: "ffmpeg
> -i
> > bbb_2.ts -c copy -f mp4 bbb_2_remuxed.mp4". If you open it up in
> Quicktime
> > Player or Safari, the first frame IS black. I think I traced it down to
> an
> > extra entry in the "elst" atom. If you probe this remuxed mp4 with trace
> > level debugging on ("ffprobe -v trace bbb_2_remuxed.mp4") there's a short
> > duration entry in the edit list of the video track.
> >
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fd357104bc0] type:'elst' parent:'edts' sz:
> 40
> > 8 40
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fd357104bc0] track[0].edit_count = 2
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fd357104bc0] duration=15 time=-1
> rate=1.000000
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fd357104bc0] duration=10000 time=2999
> > rate=1.000000
> >
> > According to the Quicktime file format docs, the parent of the "elst"
> atom
> > is "edts" and it is optional, so I removed that atom using mp4edit in the
> > Bento4 tool kit. This produced "bbb_2_remuxed_no_edts.mp4". If you open
> > this file in Quicktime or Safari, the first frame is NOT black.
> >
> > bbb_2_transcoded.mp4 is a transcoded version using "ffmpeg -i bbb_2.ts
> > -video_track_timescale 90K -movflags +faststart bbb_2_transcoded.mp4". If
> > you open this file in Quicktime or Safari, the first frame is NOT black.
> If
> > you inspect the atoms in this file, the video "elst" atom is still there,
> > but it only has a single entry in the list.
> >
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fe3ecf06040] type:'elst' parent:'edts' sz:
> 28
> > 8 28
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fe3ecf06040] track[0].edit_count = 1
> > [mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fe3ecf06040] duration=10017 time=3000
> > rate=1.000000
> >
> > So, ultimately I think the extra entry in the elst edit table is the
> cause
> > of the issue. My question is whether it is possible to not have that
> happen
> > via the ffmpeg/libav invocation, or whether this is a problem with the
> > source material?
>
> Reset the pts i.e. first packet of video stream should have pts 0.
>
> Regards,
> Gyan
>
> _______________________________________________
> ffmpeg-user mailing list
> ffmpeg-user at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-user
>
> To unsubscribe, visit link above, or email
> ffmpeg-user-request at ffmpeg.org with subject "unsubscribe".
>