Dictate and listen

Two different things, in opposite directions: dictating turns your voice into text in the composer; listening reads an answer of the assistant aloud.

Dictate a message

1. In the composer, press and hold the microphone.
2. Speak. The text appears while you talk, so you know it is following you.
3. Release. The text stays in the field: read it and correct it before you send.

Recognition is set to Italian in this release. The service may stop by itself after a long silence, even if you have not released; the indicator then stops saying it is listening.

Dictation is not offline. It uses the speech recognition of the browser engine, which sends the audio to an online service: it runs inside the window, not inside this computer. There is no wake word and no always-on listening, and a window that loses focus also loses the microphone.

Listen to an answer

A button on an answer of the assistant reads it aloud, with the voices installed in your operating system. Reading aloud is local.

If something goes wrong

Each error has its own sentence:

• Microphone permission denied — allow the microphone for TALOS in the system or browser settings.
• No speech detected — no audio arrived.
• The speech service cannot be reached — it is the network.
• No microphone found — there is no device.
• Voice not available — this browser engine does not offer speech recognition at all.