DOCS
Features

Voice

Let players speak to your game — speech-to-text as a platform primitive, and free.

A game can take spoken input. The player holds a button, talks, and your game gets text. It costs nothing — the recognition is the device's or the browser's own, not a service we bill you for.

Asking for it

Let the player hold a button and say a command instead of typing it.
The player should be able to talk to the shopkeeper out loud.

What the game gets while the player speaks:

  • A live partial transcript, updating as they talk — worth showing, since it's how a player knows the microphone is actually working.
  • A microphone level, 0 to 1, so you can drive a waveform or a glowing ring.
  • The final text, once.
  • A reason when it fails — permission denied, no speech, no network.

Turning it on

Two things, both one-time:

  1. Declare the microphone capability in Game Settings. Like the AI capability, a game that hasn't declared it doesn't get it — the browser itself refuses.
  2. The player grants permission, in their browser or OS, the first time.

Always give players a keyboard

Voice is not available everywhere. Some browsers don't implement it, some players deny the permission, and some are in a room where talking is not an option. The platform tells your game whether voice is supported, so the right design is always voice and typing, never voice only:

Voice input for the command box, with a text field as the fallback.

Pick the language deliberately

Recognition does not guess the language. You have to tell it which one you expect, and getting this wrong looks exactly like a broken microphone — the player speaks Chinese and gets confident nonsense in English.

The sane default is the player's own language, which the platform hands your game at startup:

Use the player's language for voice input, and let them change it in settings.

See Languages.

What it is not

This is speech to text, not voice chat and not a voice for your characters. There is no text-to-speech and no player-to-player audio. If you want an NPC to answer what the player said, feed the transcript to a living NPC — voice in, text out, and the player reads the reply.

Next steps

On this page