ZONOS2 Technical Report

About

We present ZONOS2 8B, our latest TTS model, which achieves state-of-the-art naturalness, prosody, and voice cloning fidelity. We improve upon Zonos-v0.1 across scale, data, and training recipe. We scale the model from 1.6B to 8B total parameters (900M active) with a novel mixture-of-experts (MoE) backbone, improving inference latency and throughput. We expand our training corpus from 200K to over 6M hours using a new data processing pipeline, and we simplify our post-training and conditioning recipes to improve naturalness and voice cloning fidelity. We evaluate ZONOS2 8B on quality, speaker similarity, WER, and ZTTS1-Eval, our novel TTS benchmark, where it performs competitively with state-of-the-art systems while maintaining good streaming latency. We release our model weights and example inference code under an Apache 2.0 license on GitHub and Hugging Face.

Gabriel Clark, Sofian Mejjoute, Mohamed Osman, George Close, Beren Millidge · 2026

Related benchmarks

Task	Dataset	Result
Text-to-Speech	ZTTS1-Eval In-The-Wild es	WER2.1	10
Text-to-Speech	ZTTS1-Eval In-The-Wild fr	WER2.79	10
Text-to-Speech	ZTTS1-Eval In-The-Wild it	WER2.51	10
Text-to-Speech	ZTTS In-The-Wild pl 1 (Eval)	WER5.85	10
Text-to-Speech	ZTTS1-Eval In-The-Wild (pt)	WER2.94	10
Text-to-Speech	ZTTS1-Eval In-The-Wild (ru)	WER4.32	10
Text-to-Speech	ZTTS1-Eval In-The-Wild (de)	WER3.37	10
Text-to-Speech	ZTTS1-Eval In-The-Wild ko	WER4.59	10
Text-to-Speech	ZTTS1-Eval In-The-Wild (en)	WER2.21	10
Text-to-Speech	ZTTS1-Eval In-The-Wild (zh)	WER2.77	10