Skip to main content

Voice forge

The voice forge lives inside Avatar Studio. You describe a voice, hear three takes, and keep the one you like under a name. That named voice can then sit on any avatar.

What it does​

  1. Open Avatar Studio and go to the Voice node, or open the Voices view and press Create a voice.
  2. Pick a type: person, animal or robot. Adjust traits (height, pace, tones) or write free words.
  3. Start a run. The forge produces up to three takes.
  4. Listen, then keep the take you want. It enters your catalogue of made voices on this machine.

The engine is Qwen3-TTS VoiceDesign (Apache-2.0). It invents a voice from the description. It does not pull voices out of another cloud account.

Free and local​

On a machine that can run Qwen3, designing voices this way has no per-voice fee and no cloud bill. You download the design model once. You can keep as many named voices as you want on disk.

Platform note​

Qwen3-TTS (speak, clone and design) installs on Windows x64 for now. On macOS and Linux the forge says so instead of staying empty. Those platforms still use Piper, system voices and ElevenLabs for read-aloud.

Assigning a made voice​

In Avatar Studio › Voice, pick one of your made voices. The avatar wears it on every desktop that carries that avatar. You can rename or remove a made voice from the Voices library.