SupraTTS 0.1 Beta

starting…

Glow-TTS + HiFi-GAN (LJSpeech voice, ~44M params) running entirely in your browser on WebGPU. Audio streams sentence by sentence.

Fetching model…

WAV

Ready when the model is.

Controls

Speed 1.00× Expressiveness 0.333 Rhythm variation 0.80 Sentence pause 120 ms Seed random Weights

Single speaker (LJSpeech). The model has no voice cloning, emotion or style controls — its knobs are speed, the two noise scales and the random seed. Same seed + settings = same audio.