Voice mode
Hands-free call
Section titled “Hands-free call”Click the Voice mode button (the sound-wave icon beside the microphone) to start a call. It’s available on desktop with a model selected.
A full-screen orb appears and pulses with your voice. Just talk:
- Listening… — speak naturally. When you pause for a couple of seconds, CoChat takes what you said as your message.
- Thinking… — the assistant works on a reply.
- The reply is spoken back to you as it’s written. Tap the orb to interrupt and speak again.
Mute pauses your microphone. End call or Esc hangs up.
Everything said in a call is a normal chat message: it’s saved in the conversation and billed like a typed one. The call keeps going when your first message turns the home page into a new chat.
If you want to be able to talk over the assistant while it’s speaking, turn on Allow Voice Interruption in Call in Settings → Interface.
Dictate a message
Section titled “Dictate a message”Click the microphone in the message box (it’s there whenever the box is empty). It turns red and a waveform shows it’s listening. Click again to stop; the transcript is added to your draft so you can edit it before sending.
Set your language in Settings → Audio → Language (for example en, fr, de) for better accuracy.
Read a reply aloud
Section titled “Read a reply aloud”On any assistant reply, open the More menu (⋯) and choose Read aloud. Choose Stop reading to stop.
Audio settings
Section titled “Audio settings”Your profile menu → Settings → Audio. Click Save after changing these.
| Setting | What it does |
|---|---|
| Speech-to-Text Engine | Default uses CoChat’s transcription. Web API uses your browser’s own speech recognition, which runs on your device. |
| Language | The language you speak, as a short code such as en. |
| Text-to-Speech Engine | Default uses CoChat’s voices; Web API uses your browser’s built-in voices. |
| Voice | The voice to use. Leave blank for the default. |
| Playback Rate | Speed of read-aloud and call replies. 1 is normal. |
If you hear a robotic browser voice instead of a natural one, check that the text-to-speech engine is set to Default.