Zonus3-Mini: Compact Zero-Shot Voice-Cloning TTS Model

Zyphra has announced ZONOS3-Mini, an open-source real-time text-to-speech voice cloning model with 800 million total parameters. The model utilizes a sparse Mixture of Experts architecture and a custom audio codec to achieve strong performance, lower word error rates, and extensive voice control capabilities including accent, age, gender, and emotion steering. Open article →