ONE voice
Voice-to-text for VS Code powered by OpenAI Whisper API with optional GPT grammar cleanup.
Press a key, speak, press again — your words appear at the cursor, in your clipboard, or directly in the terminal. Same Whisper model that powers ChatGPT voice input, with optional grammar post-processing that fixes punctuation, filler words, repeated words, and broken sentences while preserving your original wording.
Requirements
- sox audio tool:
sudo apt install sox (Linux) or brew install sox (macOS)
- OpenAI API key with Whisper access
Setup
- Install sox (see above)
- Set your OpenAI API key:
- VS Code Settings:
one-voice.apiKey
- Or environment variable:
OPENAI_API_KEY
Usage
- Open any text file
- Press Ctrl+Shift+R (or Pause/Break) to start recording
- Speak
- Press Ctrl+Shift+R (or Pause/Break) to stop and transcribe
- Text appears at your cursor / clipboard / terminal (configurable)
Keybindings are fully customizable via VS Code Keyboard Shortcuts.
Output Targets
Choose where transcribed text goes (one-voice.outputTarget):
- clipboard (default) — copies to clipboard, paste anywhere with Ctrl+V
- editor — inserts at cursor in the active file
- terminal — types into the active terminal (for Claude Code, Copilot, or any CLI)
Grammar Cleanup
When enabled (one-voice.grammarCleanup), transcriptions are post-processed by GPT to:
- Fix grammar, punctuation, and capitalization
- Remove filler words (um, uh, like, you know)
- Remove repeated words and stutters
- Construct complete sentences from fragmented speech
- Correct misheard brand names (e.g., "chachapiti" → "ChatGPT")
The grammar model, prompt, and behavior are fully configurable.
Extension Settings
| Setting |
Default |
Description |
one-voice.apiKey |
"" |
OpenAI API key |
one-voice.model |
gpt-4o-transcribe |
Whisper model (gpt-4o-transcribe or whisper-1) |
one-voice.language |
en |
Language code (ISO 639-1) |
one-voice.outputTarget |
clipboard |
Output destination: clipboard, editor, or terminal |
one-voice.grammarCleanup |
false |
Post-process with GPT for grammar cleanup |
one-voice.grammarModel |
gpt-4o-mini |
GPT model for grammar cleanup |
one-voice.grammarPrompt |
(built-in) |
Custom system prompt for grammar processing |
one-voice.maxDuration |
300 |
Max recording duration in seconds |
one-voice.soxPath |
sox |
Path to sox binary |
Install
From VS Code Marketplace:
Search "ONE voice" in the Extensions panel, or install from the marketplace page.
From GitHub (manual):
# Download the latest .vsix from GitHub releases
code --install-extension one-voice-x.x.x.vsix
Status Bar
- $(mic) ONE voice — idle, ready to record (green text)
- $(record) ONE voice 1:23 — recording with elapsed timer
- $(sync~spin) ONE voice transcribing... — processing audio
License
MIT
ONE voice is part of the ONE Ecosystem // powered by EXPL Nodes
