Zyphra open-sources ZONOS3-Mini: 800M-param voice-cloning TTS that runs in your browser
victormustar · x · 2026-11-10
Zyphra released ZONOS3-Mini, an Apache 2.0 open-source TTS model with 800M total parameters (100M active) offering natural speech and high-fidelity voice cloning. It runs in real time on phones and laptops—including fully in-browser via WebGPU with no install or server. It clones voices from an audio sample or designs characters by age, accent and texture, with per-word control of emotion, pace and delivery. Deployment options include the browser-based Zonos Voice Studio, an OpenAI-compatible /v1/audio/speech endpoint via SGLang-Omni (streaming, transcript-free cloning), and Zyphra Cloud.
Related event: Zyphra Open-Sources Zonos3-Mini: On-Device Real-Time Voice Cloning TTS(6 posts)→
More from Multimodal
- Open-source PersoDub dubs videos in your own voice across 100+ languages, locally — bigaiguy · 2026-11-11
- Testing MiniMax H3-Image for fashion contact sheets: multi-frame consistency holds up — Hot-Purchase-3738 · 2026-11-11
- Midjourney --sref 1109104433: painterly urban crowd illustrations — sergeantsref · 2026-11-11
- MiniMax H3-Image turns one prompt into an industrial robotic-hand engineering sheet — LearnWithBishal · 2026-11-11
- Gothic 2D character portrait prompt template with three sample renders — azed_ai · 2026-11-11
- Open-Source AI UGC Workflow Makes TikTok-Style Clips for $0.67 Each vs $50 Creators — aliscodes · 2026-11-11