Voice

X6 can listen and speak. Voice features appear once the operator has added keys for a speech-to-text service and a text-to-speech service; without them the controls stay hidden. The keys stay on the server. The browser only receives short-lived tokens that expire after about a minute.

Dictation

+ / More actions → Dictate streams your microphone to the speech-to-text service, and your words appear in the message box as you speak. Nothing is sent until you choose Send.

Read aloud

Each reply gets a Read aloud button. The voice is chosen in Settings → Voice.

Transcribe a recording

Open an audio or video file from the workspace in Preview and choose Transcribe. X6 saves <name>.transcript.md next to it, with speakers and timestamps. X6 can also transcribe a file itself during a task with its transcribe_audio tool.

Talk: hands-free conversation

With both services set up, + / More actions → Conversation mode starts a spoken conversation in the current chat.

1. Speak, then pause. When you stop talking, your words are sent.
2. Listen. When X6 finishes, its answer is read aloud. While Talk is on, X6 keeps its answers short.
3. Interrupt any time. Start talking over the answer and X6 stops to listen.
4. End it with Esc or the Conversation mode button.

If a task needs your approval, X6 says so out loud.

Tip

Headphones give the cleanest interruptions. On speakers, Talk relies on the browser’s echo cancellation.

Microphone access

Browsers only allow the microphone on HTTPS pages or on the local address, so dictation and Talk work on the secure addresses and on the computer running X6.