MossTTS is incredible, please port to Metal (Apple Silicon)

[vdruts]

I've tried ALL the models; and am floored by Moss TTS's realism, and speed in cloning. It's especially good at long-form generations. I use it in Openclaw for personal agents. It would be incredible to be able to use it on M2-M4 Apple Mini's etc, but right now it's only CUDA :/

[blazingbhavneek]

@bbecausereasonss Open moss team is working to make sglang as inference backend, and sglang may support Apple (Some recent issues have proposed Apple hardware support) so it might take some time but its possibility

Or you can patch current inference to use MLX as backend for Qwen backbone?

[YWMditto]

The official quantitative version and llama.cpp are both open source. Please check the latest readme for usage instructions.

[YWMditto]

A small model that can run on a CPU and has excellent voice cloning capabilities has been officially added to our R&D plan. Please stay tuned!

[YWMditto]

If there are no further questions, we will close this issue. Please feel free to continue asking any questions; we will answer them very patiently. Thank you for using and following our model!