Modbot Docs
Companion

Dictate

Speak a question to AI Chat from the headset panel or the desktop overlay. Your words are written out on your PC, sent to AI Chat, and the answer shows in a Chat panel. Your voice is never saved or sent.

In a headset you have no keyboard. Dictate lets you ask AI Chat out loud instead of typing: tap a button, say your question, tap again, and the answer appears in your headset. On a monitor it works the same way from the desktop overlay.

Using it

  1. Tap Dictate on the small button panel, under Hide or Show. The button reads Listening and a row of bars rises and falls with your voice.
  2. Say your question.
  3. Tap the button again. It reads Writing while your words are written out as text, then Sending while they go to AI Chat.
  4. The Chat panel opens as soon as your words are sent. It shows them with Waiting for an answer under them, and the answer appears there as it is written. The button goes back to Dictate.

To ask a follow-up, tap Dictate again. The new question goes into the same conversation.

The conversation is the same one you see on the website's Chat page, so you can carry on there with a keyboard, or read it again later.

If nothing was said, or nothing could be made out, nothing is sent and the button says Nothing heard for a few seconds, then goes back to Dictate.

When something fails the button says so for a few seconds: Could not listen, Could not write, Could not get ready or Could not send. A failed send also shows its reason in the Chat panel.

On the desktop

The desktop overlay has a Dictate button in the strip along its top, next to Close. It works the way the headset button does: click it, say your question, click it again. The button shows the same words, Listening, Writing, Sending, and the same bars. The answer shows in a Chat window.

You can also start and stop Dictate with a key, which is useful while VRChat has the keyboard. The key is set in the app's Settings, on the Desktop overlay card, in the Dictate row: press Record and then the keys you want. Clear takes it away again. There is none until you record one, and it works while the desktop overlay is switched on. Press it once to start and again to stop, just like the button.

When the button is there

The Dictate button, on the headset panel and on the desktop overlay, only shows when both of these are true:

  • AI Chat is switched on for your server (Chat).
  • You are allowed to use it: your role has Use AI chat (roles and permissions).

If either stops being true, the button goes away. The companion checks when it starts, every five minutes, and when you pair it with a different server.

The Chat panel

The answer shows in its own panel, in your headset or in a window on your monitor. It is held the same way as the other panels:

  • Lock stops it being picked up, moved or resized.
  • Hand makes it click-through, so pointing and clicking go to VRChat behind it.
  • You can pick it up and move it, and drag its corner to make it bigger or smaller.
  • The X closes it.

In the headset, where you put it is remembered, and two quick grips send it back to its own spot. See the bar under the panel for what the buttons do.

On the desktop the Chat panel is its own window. It opens on the screen VRChat is on, in front of it, and never off the screen. Drag its top strip to move it and its bottom-right corner to resize it. The Lock and the Hand are in the strip when Edit mode is on, the same switch as for the other panels. Its place and size are remembered, and the X closes it until the next time Dictate has an answer to show.

The first time

The first tap downloads a speech model of about 108 MB, once. The button reads Getting ready with how much has come, as a percentage. Nothing you say is heard until it is done. When it is done the button says Ready, then goes back to Dictate, and you tap it again to start speaking, so you are never recorded the moment a download ends.

Dictate understands English only.

What happens to your voice

  • Your voice is turned into text on your own PC. It is not sent to a speech service.
  • The sound is never saved. It is held only until your words are written out, then thrown away.
  • Only the written words are sent, to AI Chat on your own server, the same as if you had typed them.
  • The microphone is shared the way a voice chat program shares it, so VRChat and Discord keep working. It is the one you picked for Listening, or the Windows default.
  • The microphone is only open between your two taps.

The model comes from GitHub, once, and nothing about you goes with the request. The privacy policy says the same.

How long you can speak

A recording stops by itself after 60 seconds, as if you had tapped again.

Modbot is not endorsed by or affiliated with VRChat Inc. or Discord Inc. VRChat is a trademark of VRChat Inc. Discord is a trademark of Discord Inc. SteamVR is a trademark of Valve Corporation. Windows is a trademark of Microsoft Corporation. Docker is a trademark of Docker, Inc. PostgreSQL is a trademark of the PostgreSQL Community Association of Canada. Railway is a trademark of Railway Corp. Other names are trademarks of their respective owners; see Legal. Modbot and its logo are trademarks of the Modbot project. Open source under AGPL-3.0.

On this page