ai subtitles
requirements
- ffmpeg
- mpv
- yt-dlp (if streaming video)
- whisper.cpp
- pre-compiled binaries for windows users can be found here
- if you have an nvidia gpu I highly recommend using the cublas version (requires cuda)
- compiling with vulkan support is also possible as a cross-vendor solution
- whisper model
- medium is supposed to be the best for non-english transcription and runs fairly quick even on older hardware
- users with higher-end hardware may also want to try out large-v3
- whisper.lua (haven't tested on windows, but it should work fine)
method
- identify the appropriate
mpv
config location for your operating system
- create a directory called
scripts
in the config location and save whisper.lua
to it
- open
whisper.lua
in a text editor and locate the WHISPER_CMD
variable
- edit the variable so that the executable name matches the one you downloaded
- edit the variable so that the
-m
option points to the whisper model file you downloaded
- open a video with
mpv
, pause it and wait for the cache to fill up to 20-30 seconds (use i to view the cache status)
- unpause and use shift+w to run
whisper.lua