Claude Voice Input

Voice-to-text for VS Code. Press a button, speak, get text.

Uses OpenAI's latest gpt-4o-mini-transcribe model — handles mixed languages in one sentence beautifully:

"Hey, давай обсудим этот feature, бо я думаю що треба переписати цей component"

Russian, English, Ukrainian, or any mix — it just works.

How it works

1. Click Голос in the status bar (or Ctrl+Alt+R)
2. Speak
3. Click again to stop
4. Text is copied to clipboard + Claude Code input is focused
5. Press Ctrl+V to paste

Setup

pip install sounddevice numpy scipy openai

Set your OpenAI API key (pick one):

# Windows PowerShell
[Environment]::SetEnvironmentVariable("OPENAI_API_KEY", "sk-...", "User")

# macOS / Linux
echo 'export OPENAI_API_KEY="sk-..."' >> ~/.bashrc && source ~/.bashrc

# Or just create .env file in your workspace root
OPENAI_API_KEY=sk-...

Restart VS Code. Done.

Cost

~$0.006 per minute of audio. A typical 3-second voice command costs $0.0003. Even 100 uses per day = ~$0.90/month.