whisper - Local Speech-to-Text & Subtitles

The whisper module provides a high-performance local speech recognition capability using whisper.cpp. It handles everything from model management to video subtitle merging.

When to Activate

• When the user wants to transcribe an audio file into text.
• When generating .srt subtitle files from audio/video.
• When merging generated subtitles into a video file.
• When performing real-time speech-to-text using LiveKit or Streaming.

Core Principles & Rules

• Local Processing: Emphasize that transcription happens locally without uploading data.
• Model Selection: Allow users to choose from different model sizes (tiny, base, small, medium, large) for speed vs. accuracy.
• File Integrity: Ensure input audio files are accessible.

Additional Scenarios

• SRT Generation: Use dictate --srt to create industry-standard subtitle files.
• Video Integration: Use merge to embed subtitles into a video stream.