Sliding A/V sync with XDCAM HD encoder

[crispyjones]

Running into lipsync issues with the XDCAM HD encoder processor and certain mp4 files. I believe the problem files have come in from iOS devices.

Symptom is that the overall length of the clip is longer post encoding by FFAStrans. The source mp4 file is 1045.33 seconds and my output mxf file is 1046.479, a second and one frame(ish) longer than the length of the original clip. This manifests itself as the video increasingly lagging the audio (or audio leading video) the longer the clip plays. I've tried various settings within the XDCAM HD encoder, various bitrates, framerates, bitdepths, rasters, and GOP all with no change. Looking at the info in webinterface both mediainfo and ffprobe all seem to agree on the input duration of the file, but that's not what I get on the output. I did try the workflow with a DNX encoder and the output file was the correct length (1045.33 in and out) with no issues.

I pared down the testing workflow to just a watch folder, XDCAM HD encoder, and delivery so I don't think that's worth posting.

I can share the problem source file, I'd prefer to do it in a DM if possible. And don't get excited I can assure you it's very boring.

[admin]

Hi cristpyjones

What is worth posting is the log for the job. Can you do that? Also, you can DM me with some download link for the source file. But my first unch here is some gop anomalie in the XDCAM encoder, which is probably not FFAStrans related per se. but rather ffmpeg. However, if I get the source file we will probably be able to check it out.

-steinar

[crispyjones]

DM sent. Yes, ffmpeg is the guts of the thing so that is the most likely culprit.

[admin]

Acually, I think it might be related to some premature changes in ffastrans but I'm not 100%. I will need to give you a private build component in order to test it. I'll send it via DM later today.

-steinar

[crispyjones]

That worked. Only tested with that one file on my test machine. I will probably hold off until tomorrow putting it on the production machine so I can keep an eye on it.

[crispyjones]

The updated processors.a3x file has been running just fine for the last two days on the production machine. No reports of issues so far.