whis-0.6.1 is not a library.
Introduction
The terminal-native voice-to-text tool. Record, transcribe, paste — all from your shell. Supports hotkey mode, presets, and pipes nicely with AI assistants.
Quick Start
Usage
# Record once
# Hotkey mode (background)
# From file or stdin
# Output options
# Presets
# Post-process with LLM
# Configuration
Environment Variables
API keys can be set via environment variables instead of whis setup:
OPENAI_API_KEY=sk-...
MISTRAL_API_KEY=...
GROQ_API_KEY=gsk_...
DEEPGRAM_API_KEY=...
ELEVENLABS_API_KEY=...
OLLAMA_URL=http://localhost:11434 # Default
OLLAMA_MODEL=qwen2.5:1.5b # Default post-processing model
Requirements
- API key from OpenAI, Mistral, Groq, Deepgram, or ElevenLabs — or use local Whisper (no API key needed)
- FFmpeg (
sudo apt install ffmpegorbrew install ffmpeg) - Linux (X11/Wayland), macOS, or Windows
For hotkey mode (one-time setup on Linux):
|
&&
# Logout and login again
Prefer a GUI?
See whis-desktop — same functionality, with system tray.
License
MIT