Voice Input and Shortcuts

Cola's voice input is useful for quickly briefing a task. You can say what to do, where to look, and what result you want. Cola transcribes speech locally, then sends the text like a normal typed task.

Start Recording

By default, when the Cola app is focused, press Option once to start recording and press Option again to stop and send. You can also click the microphone button in the main interface.

If macOS asks for microphone permission the first time you use voice input, allow it. Cola still works with typed input if you do not allow microphone access, but voice input will not work.

The macOS microphone permission is usually here:

System Settings > Privacy & Security > Microphone

After confirming Cola is allowed to use the microphone, reopen Cola and try again.

Say the Task Clearly

Voice tasks do not need to sound like rigid commands, but it helps to include three things:

• Goal: what you want Cola to do.
• Scope: which file, directory, webpage, attachment, or project it should use.
• Output: whether you want a summary, checklist, edit, document, or file.

For example:

Look at these meeting notes and turn them into an action list, with an owner and due date for each item.

Check the screenshot I just attached, tell me the most obvious usability issue, and do not edit any files.

Read this project's documentation, summarize the structure first, then tell me where to look next.

For sensitive tasks, ask Cola to plan first:

First tell me which files you plan to read, then wait for my confirmation.

Fix a Bad Transcript

If the transcript does not match what you meant, the safest fix is to correct it immediately with text:

I misspoke. Limit the scope to PDFs in Downloads and do not inspect other files.

In the previous instruction, change "delete" to "list". Do not delete anything.

You can also record a shorter task again. For complex work, split the brief into two steps: first state the goal, then add file scope and output format.

Adjust Shortcuts

Open Settings > Shortcuts to adjust:

• Voice Shortcut: start and stop recording.
• Wake Mini Chat: open Mini Chat globally.
• Send Message: choose Enter, Command + Enter, or another available send shortcut.

The Voice Shortcut works inside Cola when the app is focused. The Mini Chat shortcut is global, so choose a combination that does not conflict with macOS, input methods, or apps you use often.

If Settings says a shortcut is already in use, choose another combination and save again. Shortcuts such as Option + Space and Ctrl + Space are commonly reserved by input-source switching and usually should not be used for Cola.

Choose Microphone and Speaker

Open Settings > Audio Devices to choose:

• Microphone: the input device Cola records from.
• Speaker: the output device for Read aloud and voice previews.

If you use Bluetooth headphones, recognition quality can be affected by call mode. If recognition is poor, try the Mac's built-in microphone or a wired microphone.

Common Causes of Poor Recognition

Check in this order:

1. Move to a quieter environment.
2. Get closer to the microphone and start with a short sentence.
3. Confirm the microphone device in Settings.
4. Confirm that no meeting app, recording app, or browser tab is occupying the microphone.
5. Restart Cola and try again.

Cola's speech recognition runs locally. A slow network usually does not directly affect recognition quality, but it can affect task submission, model replies, and external service calls after transcription.

Privacy Notes

The microphone is used only while you actively record. The transcribed text becomes task content for Cola. If the task needs a model or cloud service, the relevant text and context may be sent with the task.

Do not speak passwords, verification codes, API keys, private keys, or card details into Cola. For sensitive work, type the scope and limits explicitly.

For more data boundaries, see Privacy and Local Data. If the voice button does nothing, see Troubleshooting.