Hi Marton,
‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Thursday, 30 de May de 2019 20:48, Reimar Döffinger <reimar.doeffin...@gmx.de> wrote: > On 28.05.2019, at 12:48, Andreas Håkonandreas.ha...@protonmail.com wrote: > > > Hi, > > This patch supersedes #13105 (https://patchwork.ffmpeg.org/patch/13105/) > > A new (simpler and more robust) implementationof of the reinsertion of the > > missing headers in the MPEG-2 bitstream from the HW QSV encoder. > > The problem is quite simple: The bitstream generated by the MPEG-2 QSV > > encoder only incorporates the SEQ_START_CODE and EXT_START_CODE > > headers in the first GOP. This generates a result that is not suitable for > > streaming or broadcasting. > > With this patch the "mpeg2_qsv" encoder is at the same level as the > > "mpeg2video", as the software implementation repeats these headers by > > default > > in each GOP. > > I am sure I might be missing things, but I would think that this is only > necessary > when muxing into certain containers. > I.e. I would think when muxing the stream into .mov, .mkv or .nut files your > patch > would only needlessly bloat the size. Yes and no. First of all, please let me know which MPEG-2 encoders do not repeat these headers when using MPEG-TS. For streaming repetion is absolutely required. And I feel that's the reason why the native mpeg2 encoder does it. > I admit this also applies to the native mpeg2 encoder, so this might be a > design issue > in FFmpeg, but it's the background why it feels a bit wrong to do it in the > encoder... Take note that the "dump_extra" filter with the parameter "remove" can be used to remove the repetition. But contrary to popular belief (IMHO), there is no way to repeat these headers using a bitstream filter. So for this reason this patch is necessary. In any case, if you suggest that this behaviour should be optional, then I can add a parameter to enable it and leave it disabled by default. You agree? Regards. A.H. --- _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
