New: ZONOS3-Mini clones your voice locally in 43 languages (Apache 2.0). Here’s how it compares with ElevenLabs
retained text
Zyphra released ZONOS3-Mini (\~800M params, \~100M active, 43 languages, 44.1 kHz). I cloned my voice from a \~20 s clip and ran the same 8 lines through it and through my ElevenLabs v4 clone. Here's the tutorial on how the workflow works: [https://www.youtube.com/watch?v=piPdyPqg7XE](https://www.youtube.com/watch?v=piPdyPqg7XE) I also wrote a free ComfyUI node pack that runs Zyphra’s local server (Windows + NVIDIA sets itself up, \~2 GB once; Mac/Linux point it at Zyphra’s server): [https://github.com/claudianeodev/ComfyUI-ZONOS3](https://github.com/claudianeodev/ComfyUI-ZONOS3) Disclosure: I work on Promptus. The node pack works in plain ComfyUI.
document fetched 2026-11-18 00:45 · canonical https://reddit.com/r/comfyui/comments/1xyu7q8/new_zonos3mini_clones_your_voice_locally_in_43