yadif placement after -i results in timing errors

[dpury]

When using yadif after the input file, pts errors are encountered. On long files this results in a/v sync issues. If I move the yadif filter to before the input file, the errors are not present. Output format doesn't seem to matter; I've seen it when making MPEG-2 PS and TS files, as well as MP4 files. As a workaround, I've been using yadif before the input, and another -filter:v after the input to do scale and setdar.

Report files attached.

This command
ffmpeg -report -i C:\ffmpeg\bin\bip_3s.mxf -filter:v yadif=0 C:\ffmpeg\bin\bip_3s.mp4

produces errors like this:
PTS 14, next:440000 invalid dropping st:0
PTS 17, next:560000 invalid dropping st:0
PTS 20, next:680000 invalid dropping st:0
PTS 23, next:800000 invalid dropping st:0

While this command
ffmpeg -report -filter:v yadif=0 -i C:\ffmpeg\bin\bip_3s.mxf -y C:\ffmpeg\bin\bip_3s.mp4

produces no errors.

Thank you,

dpury

[dpury]

this log shows the error when using yadif after the input

[dpury]

this log shows no errors when using yadif before the input

[cehoyos]

Replying to dpury:

As a workaround, I've been using yadif before the input

This disables the filter (filter only have an affect if specified before the output file and after the input file), so I don't think it can count as a workaround.

Do I understand you correctly that if you yadif (if you place the filter between the input and the output file) the output file plays out-of-sync?
Or are you only reporting that ffmpeg is overly verbose (outputting messages) if you use yadif?

[cehoyos]

In any case, since I cannot reproduce your problem, please provide a sample of an input file that allows to reproduce the problem.

[dpury]

This is the mxf file I used to make the reports.

[dpury]

Here's the log of an interlaced HDV file source being converted to mp4

[dpury]

Here's the same HDV file converted to mp4 without yadif.

[dpury]

Thanks for looking into it, Carl. I'll dig around for another source file that I'm allowed to share and is a reasonable size.

[dpury]

Yes, longer encodes such as full-length movies drift out of sync. I'm not worried about the verbosity.

[cehoyos]

I tested with the sample you uploaded - bip_3s.mxf - and I can neither reproduce any suspicious messages on the console nor any A/V-desync.

Please reopen if you can provide a sample that allows to reproduce your problem.

[dpury]

Please try this file. If you still can't reproduce the errors, what's the next step?

 http://www.filehosting.org/file/details/398267/motion_525c_RGB235.avi

[cehoyos]

Replying to dpury:

 http://www.filehosting.org/file/details/398267/motion_525c_RGB235.avi

A hosting site that requires registration is absolutely unacceptable, sorry.

[dpury]

Replying to cehoyos:

Replying to dpury:

 http://www.filehosting.org/file/details/398267/motion_525c_RGB235.avi

A hosting site that requires registration is absolutely unacceptable, sorry.

Sorry about that. I'll upload the file to the hosting company of your choice. Thank you.

[rogerdpack]

http://ffmpeg.org/bugreports.html describes the ftp server...

[cehoyos]

You can either upload your sample to our incoming directory (there is no enforced size limit) or upload it to any hosting site that does not require registration.

[dpury]

File "motion_525c_RGB235.avi" uploaded to /incoming. Thanks!

[cehoyos]

You originally wrote:

On long files this results in a/v sync issues.

The file you uploaded is video-only, how is it supposed to show A/V sync issues?

[dpury]

I uploaded it because the invalid dropping messages were present if I used yadif.

I have done a little more research, and the issue is not present in the latest 32-bit Zeranoe build. The above reports were made with the latest Zeranoe 64-bit build. I'll take the issue to the Zeranoe forum, and I apologize for wasting your time.

[cehoyos]

Replying to dpury:

I uploaded it because the invalid dropping messages were present if I used yadif.

Above you wrote:

Yes, longer encodes such as full-length movies drift out of sync. I'm not worried about the verbosity.

This may indicate an important bug, so if you want to help, please reopen and provide a sample.