Higgs-Audio-v3 TTS
The latest TTS model released by boson.ai.
It's a model across 100+ languages, with zero-shot voice cloning and emotion control. Here is a review video made by a friend from bilibili – the model shows not only high quality output, but also great emotion (and the video is quite funny because of it).
Benchmark results using a visual novel test set from the video:
Anyway, it's probably the SOTA TTS model so far. The only problem is that the model is super huge (about 10GB). With a GGML implementation and quantization, it will be able to run on mid-range or even lower-end hardware. So I hope you can take a look at it when you have time, thanks!
Higgs-Audio-v3 TTS
The latest TTS model released by boson.ai.
It's a model across 100+ languages, with zero-shot voice cloning and emotion control. Here is a review video made by a friend from bilibili – the model shows not only high quality output, but also great emotion (and the video is quite funny because of it).
Benchmark results using a visual novel test set from the video:
Anyway, it's probably the SOTA TTS model so far. The only problem is that the model is super huge (about 10GB). With a GGML implementation and quantization, it will be able to run on mid-range or even lower-end hardware. So I hope you can take a look at it when you have time, thanks!