Voice
Dictate to motita and have its answers read aloud.
The browser interface can listen and speak. Voice is off until you turn it on in Settings → Voice.
Speech input
A microphone button beside Send dictates into the message box. The text always lands in the box to be read and edited; it is never sent on its own. A hands-free mode listens for a wake phrase (by default hey motita) while the tab is open.
Who listens is your choice:
- This browser. A Whisper model the gateway downloads once and the page runs itself, on WebGPU where available and on WebAssembly otherwise. Nothing you say leaves the device. This is the private option.
- A transcription service. The gateway forwards the recording to an OpenAI-compatible transcription endpoint (OpenAI, Groq, or a local whisper server). The audio is held in memory for that request and never written to disk.
The microphone only exists on a page served over https or from localhost, so over a plain LAN address the switches are greyed out with that reason. See Network and remote access.
Reading answers aloud
- The speaker button on an answer, or Alt+L for the last one, reads it or stops it.
- Settings → Read aloud chooses when answers are read on their own (never, only after a turn you spoke, or always), plus voice, speed and length.
- Voices can be the device's own, or local neural voices the browser runs itself.
- Natural style talks about the turn like a colleague: a short line when it starts, one about what it is doing, and the gist of the answer. Word for word reads the text as written.
- Code blocks and tables are named instead of read.
With both on, Conversation mode listens again after each answer so you can talk back and forth. Optional subtitles show what is being said above the message box.