[FFmpeg-trac] #5170(undetermined:new): feature request: conforming

[FFmpeg]

#5170: feature request: conforming
-------------------------------------+-------------------------------------
             Reporter:  zhangweiwu   |                     Type:
               Status:  new          |  enhancement
            Component:               |                 Priority:  normal
  undetermined                       |                  Version:
             Keywords:               |  unspecified
             Blocking:               |               Blocked By:
Analyzed by developer:  0            |  Reproduced by developer:  0
-------------------------------------+-------------------------------------
 When a playback device can only accept 25fps and input is 24fps, people
 usually have to re-encode. But some other software provided an alternative
 way called conforming.

 Instead of re-encoding for a different framerate, the video stream can be
 copied with a different set of timestamps and duration. This is called
 conforming. The stretch of video length is usually not observable, and the
 audio is stretched at the same rate (ffmpeg's atempo filter can do that)
 and re-encoded.

 This feature useful in the case that re-encoding video damages perceived
 quality more than alterating its duration.

 It is especially useful when audio doesn't matter: e.g. when user wants
 slow-mo effect from a 240fps camera, or when user processes security
 camera output at low framerate.

 Currently for this to work, users do the following:

 1. mencoder allow users to specific framerate, overwriting the framerate
 specified in the source file. Use mencoder to conform the video:
 {{{
 > mencoder -fps 25 -o conformed.avi -ovc copy -nosound 24fps_input.mp4
 }}}

 2. use ffmpeg to set tempo:
 {{{
 > ffmpeg -i conformed.avi -i 24fps_input.mp4 -map 0 -map 1 -c:v copy -af
 "atempo=0.96" 25fps_output.mp4
 }}}

 3. delete the temp file
 {{{
 > rm conformed.avi
 }}}

 where 0.96 equals 24⁄25.

 This feature request propose to allow user do this without the need of
 mencoder.

 This can be done by making -r parameter overwrite the framerate specified
 in input file when user specifies video stream to be copied instead of re-
 encoded.

 If user specified -r when re-encoding the video, he could mean that input
 video's frames be dropped or duplicated, or that the output video should
 be encoded in the specified framerate.

 However, if user specifies -r together with '-c:v copy' for video stream
 to be copied, he unambiguously means that he intends conforming. ffmpeg
 can proceed with conforming, and, by the end of the process, if it found
 that video and audio duration in input mathces each other, but in output
 different, it can give a warning, prompting the user to google for
 solution (which is to use atempo).

--
Ticket URL: <https://trac.ffmpeg.org/ticket/5170>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker