Post
27
Blue TTS v3 is out, an open-source TTS model that runs locally. Almost everything changed since v2.
Speech is more natural and expressive, with a MOS of 4.43 out of 5.
Voice cloning has improved significantly, and the model is fast and lightweight: it runs locally at 50x real-time on a CPU, no GPU needed.
MIT license, 388 MB, 44.1 kHz output, 7 languages: English, German, Spanish, Italian, Russian, Hebrew, Yiddish.
* Code: https://github.com/maxmelichov/BlueTTS
* ONNX weights (v3): notmax123/BlueV3-onnx
* PyTorch weights (v3): notmax123/BlueV3
* Live demo: notmax123/BlueV3
* Website: https://bluetts.com
Speech is more natural and expressive, with a MOS of 4.43 out of 5.
Voice cloning has improved significantly, and the model is fast and lightweight: it runs locally at 50x real-time on a CPU, no GPU needed.
MIT license, 388 MB, 44.1 kHz output, 7 languages: English, German, Spanish, Italian, Russian, Hebrew, Yiddish.
* Code: https://github.com/maxmelichov/BlueTTS
* ONNX weights (v3): notmax123/BlueV3-onnx
* PyTorch weights (v3): notmax123/BlueV3
* Live demo: notmax123/BlueV3
* Website: https://bluetts.com