STT skill — speech-to-text (ears) via Deepgram/Whisper #8
Description
Summary
Give Kit ears — speech-to-text for voice messages and real-time audio.
Tasks
- Evaluate Deepgram Nova-3 vs local Whisper
- Build skill following openclaw conventions
- Real-time streaming for Discord voice
- Async STT for voice messages (WhatsApp, Discord)
- Integrate with existing openai-whisper skill or replace
Notes
Hybrid approach likely best: Deepgram for real-time, Whisper for offline.