[FFmpeg-user] FFMpeg problem using fieldmatch and decimate filters to IVTC

[L. Lee]

On 5/5/14, 10:42 PM, "Steve Boyer" <steveboyer85 at gmail.com> wrote:

>Hey all,
>
>Originally, I posted this to the wrong mailing list and had a request to
>execute an ffprobe command. I am reposting and including the output of the
>commands.
>
>The ffprobe.txt file is for the sample.vob and the ffprobe.4.18.txt is for
>the full, ~2GB vob.
>
>I'm trying to encode a large number of videos so I can watch them on my
>phone, but not having luck doing a proper IVTC using the fieldmatch and
>decimate filters. I have compiled from source yesterday afternoon as
>outlined on https://trac.ffmpeg.org/wiki/CompilationGuide/Ubuntu. Using
>the
>command:
>
>ffmpeg -i 4.18 -map 0:1 -map 0:3 -vf crop=704:480:10:0 -vf
>fieldmatch,decimate -c:v libx264 -preset veryfast -level 41 -crf 22 -c:a
>libfdk_aac -t 360 4.18.fieldmatch.decimate.mp4
>
>I result in a proper IVTC'ed video stream, but the video seems to go at an
>incorrect framerate (although running "ffmpeg -i 4.18.mp4" shows it at a
>23.976 fps). I did other versions (using -vf fieldmatch, -vf yadif, -vf
>pullup,fps=24000/1001, or -vf pullup -r 24000/1001, etc) and the audio
>track remains in sync with the video consistently. Adding the decimate
>filter in to the mix seems to make the video play faster than the audio
>leading to the desynchronization. I see that
>https://trac.ffmpeg.org/ticket/3019 claims this as fixed, but I'm still
>experiencing it.
>
>Samples (a snippet of the .vob cut using dd, MP4s going through various
>filter combinations, and then a 6 minute run of the encode using the
>command above to really show the AV desync) along with text output files
>are available on my dropbox at:
>
>https://www.dropbox.com/sh/v9jp4pdxzad4a7g/djUTH4OYcA
>
>If I can help in any way, let me know! A good IVTC solution is the ONLY
>thing I miss from Windows so far. I've had hit and miss results with
>pullup
>and -r and/or -fps
>

I didn¹t solve it yet, but I have one more observation. Placing
³fps=30001/1001² in front of the items of the filtergraph sent in my
previous reply (ie: ffmpeg -i Œlong.sample.vob¹  -vf ³fps=30001/1001,
fieldmatch=order=tff, fps=fps=24000/1001² -pix_fmt yuv420p -c:v libx264
-preset slower -tune film -crf 18 -maxrate 2100k -bufsize 1835k
Œlong.sample.m4v¹) appears to make the passages that were incorrectly
rendered using the previous command line render correctly, as well as to
ruin the transcoding of passages that were previously rendered correctly
(nearly all of the sample). If it¹s possible to cause ffmpeg to switch
between the two modes for appropriate passages, then successful
transcoding for the entire clip should result. However, I¹ll admit that
the circumstance seems incorrect altogether to me. Even so, if such
automatic switching between the two modes is indeed possible, I¹d
certainly like to know how to manage it.

Laine Lee