🕊️ TontaubeV1.1 — Text to Speech

Multilingual text-to-speech with zero-shot voice cloning and five style sliders. This demo runs the released 4-bit weights, TontaubeAI/TontaubeV1.1-int4; the inference server runs the same model several times faster on your own GPU, or on a CPU. Type some text, pick a voice (or upload your own recording under Advanced settings), and move the sliders to change the delivery.

Language
Voice
Style

Each slider is optional: auto leaves that choice to the model.

Speaking rate: 0 = much slower, 4 = much faster
Pauses: 0 = fewer shorter pauses, 4 = more longer pauses
Intonation: 0 = flat monotone, 5 = very expressive
Voice pitch: 0 = low deep voice, 4 = high voice
Crisp articulation: 0 = crisp staccato, 2 = normal articulation

Bundled voices are synthetic samples released by Craitech under the MIT license. Model weights are covered by the Tontaube Community Model License 1.0. Powered by TontaubeV1.

Examples