Voice
Last updated 20 October 2026
Talk to the assistant and hear its replies read aloud, in a voice that runs entirely on your Mac.
Hear replies read aloud
The speaker button in the chat drawer’s header turns reading aloud on and off. With it on, each part of a reply is spoken as soon as it arrives, so you hear the start while the assistant may still be working. Click the button while it is talking to stop it and turn reading aloud off.
When a reply is going to be spoken, the assistant is asked to answer the way a person would say it: a few natural sentences, with no lists, tables or formatting. Your own message stays exactly as you typed it.
Talk hands-free
The microphone button beside Send starts a voice conversation, and turns reading aloud on with it. Speak, then pause: MindLake turns what you said into text on your Mac, sends it, reads the reply aloud, then listens again. The button shows each stage: a microphone while the conversation is off, a waveform while it listens, dots while the assistant works, and a speaker while it reads the reply.
- The microphone is off while a reply is being read, so talking over it does nothing. The speaker button stops it.
- Silence is never sent, and each message is sent after at most 30 seconds of listening.
- Click the microphone again to end the conversation. Whatever you were saying at that moment is not sent.
Talk about a task
In Tasks, hover a task and press the waveform, choose Talk about it from its ••• menu, or right-click it and choose Talk About It. A new hands-free chat opens, already listening, that knows the task’s line and the note it is in. Ask what the task is for or what the note says around it, or ask to change or finish it: the assistant moves its date, changes its priority, owner or tags, or ticks it off by editing that one line in the note. After an edit it says in a sentence what it changed, then listens again.
It needs AI features on and a chat connection. A task in a locked note is refused, since the assistant can’t read or change locked notes.
Where your voice goes
Your speech is recorded and transcribed on your Mac by the speech model you use for recordings, so that model must be downloaded (Settings ▸ Local Models). Filler sounds (um, uh, erm and hmm) are dropped, and the recording is deleted once it has been transcribed. Only the text leaves your Mac, as an ordinary chat message to the route you connected. With a Gemma model on your Mac (Settings ▸ AI ▸ Local), the whole conversation stays on it.
MindLake listens on the microphone chosen under Input device in Settings ▸ Audio. If nothing is heard, check that MindLake has microphone access in Settings ▸ Permissions.
Choose a voice
Settings ▸ Local Voice sets how replies sound:
| Setting | What it does |
|---|---|
| Voice | Nine voices: Alana in four styles (Sleepy, Bright, Wise and Focused), Emma, Alice, Lily and George, all British, and Heart, American. The default is Alana in the Sleepy style. |
| Heat | Warmth. Towards the flame, the voice gets warmer and a little slower. |
| Speed | How fast it talks, from the tortoise to the hare. |
| Preview | Speaks the words in Preview text with your settings, so you can hear a change straight away. |