Strata PlatformDocs

Voice mode

Voice mode lets you talk to Strata in the chat you are in. You speak; what you said is sent as a message; Strata answers it in the chat the way it answers anything you type; and the answer is read aloud. Everything chat can do works by voice: your extensions and databases, files and documents, charts, approvals, web search and the rest, under the same rules your organization sets for chat.

Three models are involved. The answer comes from the chat model you picked in the model picker, as for a typed message. Two models your organization deploys through its Azure AI Foundry carry the voice: Realtime Transcription turns your speech into text, and Realtime Voice reads Strata's answer aloud. Both are entries in the Model Catalog.

Note

Voice mode is best supported on desktop Chrome and Edge. The web app also has a separate Dictate button (the microphone icon) that turns speech into text in the prompt box for you to edit and send yourself; that is dictation, not voice mode.

Prerequisites

  • Microphone permission for the browser tab.
  • Voice mode is hidden automatically if any of these are true: the Realtime Voice or Realtime Transcription model is not deployed in your organization's Azure AI Foundry, an admin has turned voice off in policy, your organization checks your messages or answers against an inline data policy (spoken words reach the AI before they can be checked), or your organization is suspended. If you do not see the voice button next to Send, it is not available for your account today — and if you contact your admin, Strata's message tells them exactly which model or setting is missing.

Starting voice mode

  1. Click the voice mode button (the waveform icon) at the right of the prompt box. With text typed in the box, the Send arrow takes its place.
  2. Strata checks that voice is available for you, then your browser asks for the microphone. Click Allow.
  3. A voice strip appears at the top of the prompt box, and the voice button shows as on. The chat stays as it is: you can scroll, read and type as usual.

The strip says what voice is doing:

  • Connecting while voice starts.
  • Listening while it waits for you, with your words appearing as you speak (Muted while your microphone is off).
  • Thinking while Strata works on what you said, which the strip shows in quotes.
  • Speaking while it reads the answer aloud.

Talking to Strata

There is no button to hold: speak, and a short pause ends what you said. It appears in the chat as your message, in quotes, with a small microphone mark beside its time.

Strata answers in the chat with everything it does for a typed message: it queries your database, searches your email and files, makes workbooks, reports and decks, shows its steps as it works, and asks before it sends an email or changes data, exactly as in chat.

The answer is read aloud as it arrives, starting with the first sentence. Code, charts, tables and file cards are shown in the chat, not read.

  • Interrupt by speaking while Strata reads: it stops at once and listens. Stop speaking (the square in the strip) does the same without saying anything.
  • Speaking again while Strata is still working on an answer sends your new message when that answer is done.
  • A sound with no words (a cough, a door) sends nothing.

Typing while voice is on

You can type and send as usual while voice is on. The answer to a typed message is read aloud too.

Mute and End

  • Mute (the microphone in the strip) turns your microphone off and keeps voice on. Click it again to unmute.
  • End (the x in the strip), or the voice button again, ends voice. Leaving the chat page or signing out ends it too, and your microphone turns off with it.

What gets saved

What you say and Strata's answers are saved like any chat, in the conversation you are in, and follow the same settings: nothing is saved in incognito, or when saving is turned off in your settings. Voice keeps no separate transcript.

Usage and limits

  • Each message you speak is a chat message and counts toward usage like a typed one. Transcribing your speech and reading the answer aloud count as voice usage.
  • A voice session lasts at most 30 minutes; at the cap Strata says the session ended and closes it. Click the voice button to start again.
  • While voice is on, Strata keeps checking your sign-in and your organization. If your session ends, your organization is suspended, voice is turned off or a voice allowance is used up, voice closes and says why.

What can go wrong

  • No voice button next to Send. Voice is hidden because the Realtime Voice or Realtime Transcription model isn't deployed in your organization's Foundry, an admin turned voice off in policy or turned the Realtime Voice model off, your organization enforces an inline data policy on your messages or answers, or your organization is suspended. Contact your admin.
  • "Voice mode unavailable" with a usage-limit sentence. A usage limit stopped you: your own monthly limit, one of your organization's ceilings, the pool or the hourly limit, or the allowance for Realtime Voice. The voice button stays next to Send while it applies. Each time you start voice, Strata checks again and, before it asks for the microphone, says which limit stopped you and when it resets; at your own monthly limit it offers Request more usage.
  • "Voice needs the microphone." The browser was not allowed to use the microphone. Allow it in the site settings and start voice again.
  • "That was not caught. Say it again." Your words could not be transcribed. Say them again.
  • Voice stops with "The connection stopped" or "The microphone was disconnected". The network dropped or the microphone was unplugged. Start voice again.
  • Strata interrupts itself on a speakerphone. Its own voice can reach the microphone and count as you speaking. Use headphones.

Related