Groq Voice Transcribe

Fast speech-to-text for voice notes and audio files through Groq's OpenAI-compatible transcription endpoint.
Use it when you want cloud transcription via Groq instead of running Whisper locally.

Best for:

• - Telegram / Signal voice notes
• short audio clips
• Chinese, English, or mixed daily speech
• fast transcript generation for follow-up summarization or chat replies

What you need

You need a Groq API key.
Groq often provides a free developer tier / trial credits for new users.
Get one from:

• - https://console.groq.com/
• Sign in, open API Keys, then create a key

Easiest setup in OpenClaw

If OpenClaw is already running and configured, you can simply ask your assistant:

• - "Configure Groq Voice Transcribe for me"
• "Here is my Groq API key, set up Groq Voice Transcribe"

The assistant can place the key into ~/.openclaw/openclaw.json for you.

Manual setup

Set GROQ_API_KEY, or configure it in ~/.openclaw/openclaw.json under:

CODEBLOCK0

Quick start

CODEBLOCK1

Defaults:

• - Model: INLINECODE3
• Output: INLINECODE4
• Format: plain text

Common examples

CODEBLOCK2

Flags

• - --model <name>: transcription model (default whisper-large-v3-turbo)
• INLINECODE7: output file path
• INLINECODE8: hint the spoken language, for example zh, en, INLINECODE11
• INLINECODE12: optional context or spelling hint
• INLINECODE13: write verbose JSON instead of plain text

Notes

• - Audio is sent to Groq for transcription.
• This skill is meant for transcription, not text-to-speech.
• If language is omitted, Groq can usually auto-detect it, but passing --language zh often helps for Chinese voice notes.