GUIDE

VRChat text to speech for mute and non-verbal players

Many players in VRChat and on Discord do not use their own voice: some cannot speak, some are shy, some play late at night next to sleeping family. Typing into the chatbox works, but a voice gets you into the conversation.

The free SuperClonica engine turns what you type into a natural voice on your own PC, and a virtual audio cable delivers it to any app as if it were a microphone. This guide shows the setup, a tiny script that speaks what you type, and how to pick a voice that feels like yours.

Browse voices Install the SuperClonica engine

How typed speech reaches VRChat and Discord

Any app that lets you choose a microphone can take the voice this way, from games to video calls.

  • A virtual audio cable, such as the free VB-CABLE, adds a playback device, CABLE Input, and a microphone, CABLE Output.
  • Your script plays speech into CABLE Input; VRChat or Discord hears it on CABLE Output.
  • In VRChat’s audio settings, pick CABLE Output as your microphone; in Discord, set it as the input device.
  • To hear yourself, turn on Listen to this device for CABLE Output in the Windows sound settings.

Hear the voices free online

On the page of any voice, type something you would say in a world, 30 to 100 characters, choose the language and press Voice it. In a few seconds you hear whether the voice fits your avatar, with no account.

Type and speak with a small script

The engine writes out common chat shorthand, such as brb, idk, ty or gg, so you can type the way you chat.

  • Install the engine and the audio library: python -m pip install superclonica-cpu sounddevice.
  • Find the number of CABLE Input in the list of devices: python -c "import sounddevice as sd; print(sd.query_devices())".
  • Run a loop that speaks what you type, with 7 replaced by that number: tts = Superclonica(); while True: text = input('> '); audio, rate = tts.synthesize(text, 'F3', 'en'); sd.play(audio, rate, device=7); sd.wait().
  • Keep the script running: the model loads once, then a short sentence is ready in about half a second on a desktop processor.

A voice that feels like yours

Pick a voice that matches your avatar: the ten stock voices of the model are free, and the catalog has calm, cheerful, young and character voices. If you have a clean recording of your own voice from the past, 5 to 10 seconds is enough to train a voice from it in the SuperClonica studio on a computer with an NVIDIA graphics card, or you can order a custom voice on the exchange.

Good manners with a TTS voice

A text-to-speech voice is a normal sight in VRChat and on Discord, and most people are glad you can join in. Keep the volume at the level of other players, do not use a voice that imitates a real person or another player, and follow the rules of the world, server or community you are in. Character voices are fan-made: fine for fun, never for pretending to be someone.