ldenoue/Underdog-Woof-WebGPU

Underdog Woof browser weights

This is an unofficial mirror of the 31.7 MB, one-voice WebGPU model served by Underdog's browser demo at https://underdog.ai/woof. It is mirrored to let https://github.com/ldenoue/xn-ptts load the model from Hugging Face instead of shipping large binary files through GitHub Pages.

The model is Underdog Woof. Credit and licensing belong to Underdog/Conway. See the included LICENSE and NOTICE files and the official two-voice release: https://huggingface.co/ConwayResearch/Underdog-Woof.

Contents

• config.json: one-voice model configuration
• isla.safetensors: Isla voice vectors
• model_int8.parts.json: shard manifest
• model_int8.part0, model_int8.part1: int8 model weights

These files were retrieved from https://underdog.ai/woof/model/ on November 7, 2026. The browser runtime is intentionally not mirrored here because Underdog has not published a clear redistribution license for its website JavaScript.

The weight, configuration, and voice files are mirrored unchanged. Only this model card and the repository packaging were added.

Restrictions

Woof is restricted to non-commercial use unless you obtain a commercial license from Underdog. Applications must display Voice by Underdog Woof with a link to https://underdog.ai. Consult LICENSE for the complete terms.