OpenAI Whisper

GitHub Repository

Subtitle creation

To transcribe a video file into an SRT subtitle file using the medium model:

nix run nixpkgs#openai-whisper -- input.mp4 --model medium --output_format srt

High accuracy (Large models)

On a 64GB MacBook Pro M1, you can use the largest models for maximum accuracy. large-v3 is the most accurate, while turbo offers a similar accuracy at much higher speeds.

Large V3:

nix run nixpkgs#openai-whisper -- input.mp4 --model large-v3 --output_format srt

Turbo:

nix run nixpkgs#openai-whisper -- input.mp4 --model turbo --output_format srt

Available models: tiny, base, small, medium, large, large-v1, large-v2, large-v3, turbo.