arXiv:2611.16525v1 Announce Type: cross
Abstract: ZONOS3-Mini is an ultra-compact, ultra-fast, zero-shot voice-cloning TTS model. The model can run in real time on a laptop or smartphone. For its latency and size, it achieves state-of-the-art accuracy and cloning fidelity. To achieve this we propose a latent MoE backbone with only 886M parameters (100M active per token) and a new audio codec, Zyphra Audio Codec 1 (ZAC-1), which represents 44.1 kHz audio at a 15 Hz frame rate. On the new ZTTS2-Eval benchmark, ZONOS3-Mini reaches a 6.71% macro-average WER over 55 language sets (35 clean and 20 in the wild), which is competitive with larger and slower state-of-the-art models. We release the model weights, inference, and training code.
arxiv
ZONOS3-Mini: Ultra-Compact Zero-Shot Voice Cloning TTS with Open Weights
ZONOS3-Mini is an ultra-compact, real-time zero-shot voice-cloning TTS model designed to run on laptops or smartphones. It uses a latent mixture-of-experts backbone with 886M parameters (100M active per token) and a new audio codec, Zyphra Audio Codec 1, representing 44.1 kHz audio at a 15 Hz frame rate. On the ZTTS2-Eval benchmark, it achieves a 6.71% macro-average WER across 55 language sets, competitive with larger and slower models. The authors release model weights, inference code, and training code.
This is a concrete, recent open-source TTS advancement with practical edge deployment potential and competitive multilingual cloning results.
Comments & Questions (0)
No comments yet
Be the first to comment!