Voice
Talk to a pull request or your codebase and hear the answer. Choose OpenAI, ElevenLabs or your system voice, a language and a voice. Pro.
Talk to a pull request or to your whole codebase and hear the answer, or dictate a comment instead of typing it. Voice is a Pro feature. Input and output are both on by default and can be switched off separately in Settings → Voice.
#Two ways to talk
- Talk to a PR. On a pull request page, ask what the PR does and hear the walkthrough. The code it is explaining is highlighted and scrolled into view as it talks. See Understand a pull request.
- Talk to your codebase. In Analyze Project, ask about your repository and hear the answer while a graph builds on the canvas. See Understand your codebase.
Both use the same push-to-talk: hold the key the voice dock shows, ask, release to send. Replay repeats the last answer.
#Set it up
- In Settings → Voice, choose a provider: System, OpenAI or ElevenLabs. OpenAI and ElevenLabs use your own API key, saved in Settings → AI providers. The system voice needs no key.
- Grant microphone permission. On macOS: System Settings → Privacy & Security → Microphone.
- Check that Input voice (and Output voice, if you want answers read aloud) is on in Settings → Voice.
#Providers
| Provider | Notes |
|---|---|
| System | Your operating system's own voices. No key, no network, no cost. |
| OpenAI | Higher-quality voices with your own key. Choose a voice and a TTS model. |
| ElevenLabs | Natural-sounding voices with your own ElevenLabs key. |
#Dictate a comment
With Input voice on, a Dictate button appears in the comment composer. Speak, stop, and what you said is written up as a clean comment.
- Language: English (US or UK), Spanish, French, German or Japanese.
- Auto-send after pause: send when you stop talking instead of waiting for you to press send.
#Who writes the answer
The spoken answer is written by the model routed to Voice answers. See Choosing models.
#Where the audio goes
Straight from your machine to the provider you chose, using your own key, or it stays with your operating system's voice. It never touches DiffGuardian's servers.
Problems? See Troubleshooting.