Feature request: Option to prepend language code to subtitle file extension

[jonheese]

My workflow has me run faster-whisper-xxl.exe $media_dir --language en --model large-v2 --batch_recursive --skip -pp and this works great to generate subs for any new media files while ignoring the previous-generated subtitle files.

However, my media player platform infers the language of the subtitle files from the file extension and files that are just <media_file>.srt are labeled as "Unknown" language. This would be fine, except it will only default to enable subtitles on playback if the language code preset matches my default language, so having an option for Whisper to output the files as <media_file>.<lang>.srt (eg. big_buck_bunny.en.srt) would be super helpful.

I have tried just scripting a rename of the srt files after running the subtitle generation, which works fine for playback, but breaks the --skip option in step 1.

Alternatively, instead of changing the output file name pattern, it would also be acceptable to have the --skip behavior consider files with extension .<lang>.srt to be matching, so it would still skip reprocessing those media files.

I looked through the source code of faster-whisper briefly to see if I could tweak either of these places for my own custom build, but didn't see anything obvious where this could be done.

Thanks.

[Purfview]

Did you tried --postfix?

[jonheese]

Ah, I didn't know about --postfix -- thank you.

I just tried it, but it doesn't seem to work as I expected:

faster-whisper-xxl.exe "$dir" --language en --model large-v2 --batch_recursive -pp --skip --postfix .en
usage: faster-whisper-xxl.exe [-h] [--model MODEL] [--model_dir MODEL_DIR] [--device DEVICE]
...
                              audio [audio ...]
faster-whisper-xxl.exe: error: unrecognized arguments: .en

I noticed that the prefix option is listed in the usage text as [--prefix PREFIX], but postfix is just listed as [--postfix]. So I also tried not putting anything after the --postfix option, just to test, and I got:

Error: --skip and --postfix doesn't work together.

I tried looking for more documentation or examples for how the --postfix option is supposed to work, but didn't find any. Can you provide an example for how I might use the postfix option to accomplish this?

Thanks again!

[Purfview]

It doesn't take any arguments and you can't use --skip with it.

[jonheese]

Okay, then I’m completely confused. What does --postfix do exactly?

[jonheese]

🤦‍♂️ I somehow didn't realize that --help yields much better information on the cli args until just now, and there I found:

  --postfix             Adds language as a postfix to subtitle's filename. (default: False)

Okay, so that explains how it's supposed to work, but if I can't use it with --skip, I don't think it will solve my issue.

For my desired workflow, the command should only create missing subtitle files (i.e. --skip) and it should postfix each filename with the language (--postfix), both in terms of looking for an existing match and the output file to produce when necessary.

Is there a reason for these two options not working together that I'm missing?

Without this ability, I suppose I could write a script to recurse the directory, determine which files are missing and only send those to the whisper process, but it seems like the built-in options are so close to doing it the way I need.

Thanks.