Edge0 releases Audio8-TTS-Realistic model
The story so far
Edge0 has released an open-source text-to-speech model, Audio8-TTS-Realistic, which supports zero-shot voice cloning for 23 languages with a 0.6B parameter size. In the CV3 benchmark test, its average error rate is 4.86%, outperforming Qwen3-TTS-1.7B and VoxCPM2 but falling short of the 4B-level Higgs TTS 3. The model adopts a dual autoregressive architecture, including a slow branch with 24 layers and a width of 896, and a fast parallel prediction branch with 4 layers. It is trained using a learned realism critic. The weights and code are available under Apache 2.0. For developers, the model's appeal lies in its competitive multilingual accuracy and ease of fine-tuning and deployment due to its small size.
AI-written from the reports · Updated 20 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- MarkTechPostEdge0 Releases Audio8-TTS-Realistic, a 0.6B Voice Cloning TTS Model
Edge0 releases the open-source text-to-speech model Audio8-TTS-Realistic, with 0.6B parameters supporting zero-shot voice cloning for 23 languages, achieving an average error rate of 4.86% in the CV3 benchmark test, surpassing Qwen3-TTS-1.7B and VoxCPM2, but lower than the 4B-level Higgs TTS 3. The model employs a dual autoregressive architecture, including a slow branch with 24 layers and width 896, and a fast parallel prediction branch with 4 layers, trained using a learned realism critic.