still ether
Myshell published four self-developed open-source AI models, including TTS and LLM, while also collaborating on AI research and development with top-tier institutions like MIT and Princeton.
All models are available on our GitHub.
- MIT, Tsinghua University, and startup MyShell have jointly released OpenVoice, an open-source AI voice cloning model capable of rapidly generating realistic voice clones with granular control over tone, emotion, and more.
- OpenVoice leverages two AI models: a text-to-speech model and a tone converter, allowing it to mimic a user's voice with just a small audio clip and adjust the emotional expression of the generated speech.
- Read more details in this Article.
- JetMoE is a highly efficient AI model that demonstrates remarkable performance while being trained on a budget of less than $100,000.
- Despite its cost-effectiveness, JetMoE-8B surpasses the performance of Meta AI’s LLaMA2-7B model, which was developed with significantly larger financial resources.
- JetMoE proves that large language model (LLM) training can be accomplished at a fraction of the cost previously assumed, making advanced AI development more accessible.
- Read more on this [Post]
(https://x.com/MIT_CSAIL/status/1775916496503656679).
- MeloTTS is a high-quality multi-lingual text-to-speech library by MIT and MyShell.ai.
- It currently supports languages including English, Spanish, French, Chinese, Japanese and Korean.
- Read more on this Post.
