Using Onetake.

Before you start.

Public release is still being prepared; there is no official public download yet. This guide describes the current candidate version.

1. Use an Apple Silicon Mac

   System requirements for the public release are still to be confirmed. A working microphone and online speech service are required; the current internal test also requires a registered device.

2. Grant permissions yourself

   During onboarding, manually grant microphone and Accessibility permissions. This website cannot grant them for you.

3. Check your microphone and Fn key

   Complete the input-level test, introduction and Fn check. The app includes practice for chat, email, documents and mixed-language notes.

View current release status

Start where you want to write.

1. Place the cursor

   Switch to a regular text field in your app. Password and other secure fields are excluded.

2. Press Fn, then speak

   Press Fn once. Wait for preparation to finish and listening to begin, then speak. You can dictate in Chinese, English, or a mix of both.

3. Finish, then review

   Press Fn again or select Finish. After recognition and editing, Onetake attempts to write back where you started. Review and edit the text, then send or save it yourself.

Fn

Start / finish dictation

Escape

Cancel dictation without producing a result

An alternative shortcut can be configured in Settings.

Play the example

Your dictionary, managed by you.

You can manually add, edit and delete names, product names and technical terms, search your entries, or import them from a CSV file. The dictionary provides recognition hints, not a guarantee that every term will be recognized correctly.

The dictionary does not automatically learn from later edits or discover new terms.

Recovery depends on what is retained.

You can view, copy, delete or clear local history. Retention can be set to forever, a year, a month, a week, a day or no retention. If the recording is still available, you can retry recognition or export it. Retrying requires an internet connection.

What if the text was not inserted?

When insertion cannot be confirmed, check the app’s failure feedback. Where available, recover the result from the clipboard or retained local history and paste it manually. Do not assume the text was sent.

What happens if text editing fails?

If recognized text is already available, the app can fall back to that text. This does not mean new speech can be recognized offline.

What if speech is unclear or the service fails?

Missing audio, unclear speech or a service failure requires explicit failure feedback, not invented text presented as a successful result.

Can recognition be retried without the recording?

No. Retrying depends on the original local recording still being available. Do not expect history or audio to remain recoverable after deletion or when no retention is selected.

See the interrupted-insertion example

Find the right support channel.

Describe the problem with non-private content. Do not upload private recordings, transcripts, account information or other sensitive material.

Product problems

Report a product issue.

Questions and suggestions

Ask about usage or share a suggestion.

Report a security issue privately

Do not post security or privacy issues publicly.

These channels are on GitHub. A GitHub account may be required to submit.