🎤 Real-time Audio → JSON Pipeline

Whisper transcription → Gemma JSON extraction with Unsloth optimization

Transcription Settings
Transcription Engine

Choose between local Whisper or Deepgram API
Whisper Model Size

Larger models are more accurate but slower
Language

Language code for transcription
1 10

Notes:

  • Whisper: Runs locally, no API key needed. Tiny/Base are fast, Large is most accurate.
  • Deepgram: Requires API key set as environment variable: export DEEPGRAM_API_KEY=your_key
  • Buffer Duration: Shorter = more responsive but may miss context. Longer = better context but more delay.