Set up voice input

Prepare local voice input and choose how spoken conversations behave.

Before you start

Open Settings → Voice. You need a working microphone and permission for Choro to use it. Voice models are downloaded on first use or when you install them from this page.

Demo screenshot: Voice settings before model installation. Install downloads the local speech model; spoken feedback and patient turn-taking are configured below. The screen also shows the Project Talk shortcut and transcript-retention information.

Step by step

1. Read the local-model information and start installation if the required model is not present. Allow the download to finish before testing dictation.

2. Check the microphone permission when macOS prompts for it. If permission was previously denied, review Choro’s microphone access in macOS settings.

3. Choose the spoken-feedback behavior offered by Choro: Minimal, Balanced, All, or Off. Pick the level that suits whether you want a spoken discussion or primarily typed results.

4. Review turn-taking preferences before a longer Project Talk session. More patient turn-taking can help if you pause while thinking.

5. Open an agent message draft and use the Dictate with voice shortcut shown in Settings → Keyboard shortcuts (hold ⌘L by default). Try one short, non-sensitive sentence, release to insert it, and inspect the transcript before sending. Correct names, file paths, and technical terms that were misheard.

What to expect

Voice input produces editable text in the relevant workflow. Local audio processing does not remove the need to review the text before it becomes an agent request.

Troubleshooting and useful details

If nothing is transcribed, check model installation and microphone permission first. Audio is processed on the Mac and not retained; the Voice page describes transcript retention separately. Once text is sent to a coding provider, that provider handles the request according to its own setup.