Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Refresh skill docs from repo main: Bun-only install, Vosk-TTS Russian voice, OGG/Opus voice-note output, timestamp/diarization guidance, and OpenClaw audio/TTS routing.
What This Skill Does
Offline multilingual voice toolkit that transcribes audio to text, synthesizes speech, and detects languages. Supports 25 European languages for speech-to-text using NVIDIA Parakeet TDT, and English/Russian text-to-speech via Kokoro-82M and Vosk-TTS. Runs entirely on-device on Apple Silicon, Linux, and Windows with no cloud dependency.
Everything runs locally with no API keys or per-minute billing, keeping voice data off third-party servers.
When to use it
- Transcribing Telegram or WhatsApp voice messages into text
- Generating OGG/Opus voice note replies for messenger apps
- Detecting the language of an incoming audio file before processing
- Creating timestamped transcripts from meeting recordings
- Synthesizing multilingual speech locally without cloud API costs
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Transcribe this voice message and tell me what language it is in
- 1Run kesha --json on the incoming .ogg audio file
- 2Parse the JSON output for transcript text and language detection fields
- 3Read the audioLanguage and textLanguage confidence scores
- 4Return the transcript along with the detected language code
Transcript text with detected language (e.g. Russian, confidence 0.98)