ffmpeg - How to copy|extract encoding settings from existing media file?

I wished to be able do this multiple times now so I ask.

If I have an existing video or audio file, ffmpeg, mplayer and other media players can detect at least some of it's "propreties" like container, codec and bitrate used, probably various quality and encoding specific settings, etc.

How can I extract these settings from an existing file in order to use them (directly) for encoding with ffmpeg?

For example, I got an mkv video encoded with x264 with some settings, and another uncompressed avi file. I would like to "copy" the encoding settings used in for x264 mkv file to transcode the avi with the same settings.

Note: I'm looking for a way that should include no human work with "translating" the settings from a decoder to the encoder. It is okay if I need to extract the settings first and save it somewhere but I would like to be able to simply feed the read options to a command line or preset|configuration file for ffmpeg.


Solution 1:

There is no automatic way to do that. You have to look at the parameters of the original file and apply them to the output file.

In most cases, these will be the following:

  • Container format (MP4, MKV, …)
  • Video and audio codec (H.264, H.265, …)
  • Audio-specific:
    • Number of audio channels
    • Audio sampling rate
    • Audio bitrate
  • Video-specific:
    • Profile and Level (to ensure compatibility, see here)
    • Maximum bitrate limitations (e.g. for H.264)
    • Maximum video resolution, change via -filter:v scale or -s:v
    • Framerate, change via -filter:v fps -r
    • Chroma subsampling, change via -pix_fmt (e.g., -pix_fmt yuv420p should give you the best compatibility)
    • GOP size (distance between IDR-frames), set via -g
    • Other specific encoding settings

But even if you get that all right, some devices may require specific, proprietary information embedded in the bitstream.


As for the specific task of using x264, this is not going to be trivial. I'm not aware of a single script that'd take care of these tasks, which are usually done manually. For the most info about the encoding settings, on Unix/Linux or OS X, you can use mediainfo with some Bash tricks.

For example, for an x264-encoded video in an MP4 file:

mediainfo input.mp4 | grep "Encoding settings" | cut -d':' -f2- | tr '/' '\n' | sed 's/ //'

This will output a list of x264 options:

cabac=1
ref=3
deblock=1:-1:-1
analyse=0x3:0x113
me=hex
subme=7
psy=1
…

You could then manually pass these options to the x264 binary.

If you go through FFmpeg, that's a little more complicated though, as not all of x264's options can or should be mapped like this. Note that often a simple preset, tune and profile specification will do as well (as seen in x264 --fullhelp and the x264 encoding guide), and specifying the CRF level is enough.

And this is not even considering audio, where luckily, there aren't that many options.