Edge0 releases Audio8-TTS-Realistic, a 0.6B open-weight voice cloning model
AIEdge0 has released Audio8-TTS-Realistic, an open-weight zero-shot text-to-speech model with 601M main-model parameters that clones a voice from one reference clip and transcript. The model supports 23 languages under Apache 2.0, and the source reports 4.86% average CV3 error across 9 languages, versus 7.73% for Qwen3-TTS-1.7B. Performance on lower-resource languages remains weak, with 48.57% WER on Thai.
Source details
MarkTechPost · Asif Razzaq · marktechpost.comPublished · added here
Original title: Edge0 Releases Audio8-TTS-Realistic, a 0.6B Voice Cloning TTS Model