Fills the censored YouTube captions using surrounding caption context and a local audio inference model. Fully local and open-source. Watch videos on mute or in loud areas. Requested by the deaf community.
Options: - Off
Rules only Uses only context. Minimal cpu load, high precision but low coverage
Rules first (Recommended) Uses context for most fills, and the whisper model for the rest of the slots. High precision and coverage, moderate cpu usage.
Whisper first Uses whisper on every slot, with the rules filling the ones failed by the model. Maximum precision and coverage, high cpu usage.
Whisper only Uses only whisper. Maximum precision, high coverage.
