Overview
Voice input turns a longer spoken instruction into editable staged terminal text. Redock can use Apple, OpenAI, or Volcengine speech services depending on the selected provider and configuration.
The transcript is not sent immediately: you can correct filenames, technical terms, punctuation, and command fragments before giving the final prompt to a shell or coding agent.
When to use it
- Describe a multi-step implementation request when typing the full prompt on a phone would be slow.
- Explain observed behavior, expected behavior, reproduction steps, and constraints in one continuous bug report.
- Capture an idea while walking, commuting, or away from a physical keyboard, then refine it before sending.
- Dictate Chinese, English, or mixed technical instructions while preserving an editable review step.
- Add context to an agent follow-up after reviewing test output, logs, or a Web Preview result.
Dictate a complete implementation brief
Speak the objective, affected screens, constraints, edge cases, and acceptance criteria in one pass, then correct technical terms in staged input before sending it to the agent.
Report a bug while observing it
While the error is visible, describe the actual behavior, expected result, reproduction steps, device state, and recent change without switching repeatedly between the app and keyboard.
Give a fast follow-up after review
After reading logs, tests, or a Web Preview, dictate the next correction with the observed evidence and preserve the transcript for a final accuracy check.
How it works
- Choose Apple, OpenAI, or Volcengine speech in settings.
- Record the instruction and wait for transcription.
- Edit the result in staged input, then send it to the terminal or agent.
Practical guidance
Review filenames, commands, and punctuation before sending. Speech is most useful as a fast draft; staged input keeps the final instruction under your control.