Voice
Assista X has two voice features: dictation, which types your question for you, and conversation mode, a hands-free loop in which you speak and Assista X answers aloud.
What you can do
| Feature | What it does |
|---|---|
| Voice input | Speak your question. It is written into the chat box so you can check it and send it |
| Conversation mode | Speak, hear the answer, and keep talking until you end the conversation |
Voice input
- Select the microphone button at the right of the chat box ("Start voice input").
- Speak your question. The words appear in the chat box as you speak.
- Select the button again ("Stop listening") to stop, edit the text if you want, and send it.
Voice input uses your browser's built-in speech recognition. The microphone button only appears in browsers that support it, and it works on desktop and mobile browsers. If your browser blocks the microphone, you see "Microphone access denied". Allow microphone access for the site in your browser settings.
Conversation mode
- With the chat box empty, select the send button. With nothing typed it shows an audio icon and the hint "Start voice conversation".
- Allow microphone access if your browser asks.
- Speak when the overlay says Listening….
- Wait while it says Thinking…, then hear the answer while it says Speaking….
- Keep talking. Listening starts again after each answer.
- Select End conversation to leave conversation mode.
| Overlay text | Meaning |
|---|---|
| Tap to speak | Conversation mode is waiting |
| Listening… | The microphone is on |
| Thinking… | Your speech was transcribed and the answer is being prepared |
| Speaking… | The answer is being read aloud |
| Something went wrong | A step failed. Tap the overlay to start again. The loop does not restart on its own |
How listening works
- Assista X sends your speech when you stop talking for about two seconds.
- If you say nothing for about six seconds, it starts listening again without sending anything.
- A single recording is cut off at 30 seconds.
- Recordings that contain only noise are discarded.
How spoken answers differ
Answers in conversation mode are written to be heard. They are plain, conversational prose without tables, charts or lists, and they read out the information instead of showing it. Use the normal chat when you want tables and charts.
Voice models
When OpenRouter is your provider, Settings > Provider > Additional settings lets you choose a preferred speech-to-text model, text-to-speech model and voice. If you leave them empty, system defaults are used.
Tips and limits
- Speak in short, specific questions, for example "What is the total outstanding receivables?".
- Audio uploads are limited to 25 MB. A message such as "Audio file exceeds 25MB limit." means a recording was too long.
- If you see "Could not transcribe audio. Please try again.", repeat the question.
- Voice features that need an OpenRouter key can show "OpenRouter API key not provisioned for voice features." Check your provider settings.