Overview
Voice input transcribes your speech into the terminal input draft, where you can review it before sending. Different providers can be configured, including Doubao and OpenAI.
You can also use local Whisper. Download a Tiny or Base English or multilingual model in Speech Settings first. After the download, recordings are processed only on this device instead of being sent to an online transcription service.
Best used for
- Dictating longer task context or change requests for an agent.
- Describing errors, reproduction steps, or the next plan.
- Reducing keyboard input during mobile remote coding.