Modelpublic

onnx-community/KittenTTS-Micro-v0.8-ONNX

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face

2likes22downloads

Model Card

KittenTTS-Micro-v0.8-ONNX

ONNX export of KittenML/kitten-tts-micro-0.8 for browser inference with Transformers.js and ONNX Runtime Web.

Model Details


Architecture	StyleTTS 2
Parameters	40M
Precision	Int8 quantized
Sample Rate	24kHz
Runtime	WASM only (int8 quantized ops not supported on WebGPU)
Voices	Bella, Jasper, Luna, Bruno, Rosie, Hugo, Kiki, Leo

Note: This model uses int8 quantized operations (MatMulInteger, ConvInteger, DynamicQuantizeLSTM) that are not supported by the WebGPU execution provider in ONNX Runtime Web. A FP32 version is needed for WebGPU acceleration. For WebGPU support, use KittenTTS-Nano-v0.8-ONNX (15M FP32).

Demo

KittenTTS WebGPU Demo — runs entirely in your browser.

Credits

—Models by KittenML
—Original demo: KittenTTS-Demo
—Transformers.js by HuggingFace
—phonemizer.js by Xenova