[FFmpeg-user] Why does the 'detelecine' filter exist?

[Mark Filipak]

On 29/05/2025 16.47, Huspnet wrote:
>> On May 28, 2025, at 6:47 PM, Mark Filipak <markfilipak.imdb at gmail.com> wrote:
>>
>> On 28/05/2025 18.46, Alex Xu wrote:
>>> I put it there since it was in Huspnet's original command.
>>
>> Ahem!
>>
>>> I don't think it
>>> matters,
> 
> Sorry, I should have made it clear that I'm a hobbyist (in case my post didn't adequately do that for me). Also, I probably shouldn't have included "-r 24" in my example without further explanation. I have had better audio sync by using either "-r 24", "-r 23.976" to specify output frame rate, ...

Specifying output frame rate is necessary only when you want to change frame rate and don't care 
about results. I believe you are packaging as MKV? MKV has known problems preserving audio sync. 
There are many reasons why I recommend MP4.

>... but that only helps when I add a frame rate input specification to the beginning of the filter chain, such as "fps=30000/1001" ("29.97" may be accepted there instead, but I think it has not always been so). I've also tried ""-fps_mode vfr" instead of the "-r" specification so I could try "mpdecimate" after "decimate" to clean up remaining duplicates. I am becoming convinced the the duplicates that remain after applying "decimate" result from telecine errors in the source video. If that is true (that I'm transcoding video that contains flaws), then, for my most commonly transcoded sources, my best audio sync has been achieved using "-vf fps=30000/1001,fieldmatch,decimate" in the filter chain, then "-r 23.976" to specify output frame rate. But that's just my experience.

'-r' works by dropping and-or duplicating frames, only -- nothing fancy. That may be your entire 
problem.

> I really need to be working on a passage of an hour or more's duration to make a reliable assessment. Also, I worked out a pretty robust way to adjust audio tempo for cases where the audio sync isn't maintained, so I'm not left with no recourse when sync issues occur. I've done that by creating a script which asks me for a value in milliseconds for the duration by which the audio is away from the video at the end of a clip, and reminds me to precede the value by using "-" (minus) first if the audio is too short. It's an AppleScript, but the basic functions are easily decipherable for any of you who may be interested in adapting for your own scripting systems, so I'll be happy to share on request.

Please do. Sometimes a script is just a script, but sometimes, it's a good cigar.