Release plan?

[rosmur]

This is a very cool project and looks like it can fill the llama.cpp sized gap for audio models. Do you plan to scale this project to be useful for average users - creating releases that users can install? Is mac support planned?

P.S: Please consider opening the Discussions section of the repo - this post isnt technically an issue but there was no other way to raise this question

[0xShug0]

Yes I will create releases. The plan is to do this after the release of next set of music models in the coming week (Acestep, heartmula, and stable audio 3).

All models listed as "released" can run on mac. you can check https://github.com/0xShug0/audio.cpp/tree/release-0.1/examples/xcode/Qwen3TTSDemo. I didn't highlight the mac metal support because ggml support for metal is bad and many ops used in the audio models have serious performance issues. Generally ggml metal backend is not friendly to audio models. I want to optimize their performance before create a release.

[JohnClaw]

Can't wait for Windows cpu avx2 binary release to be uploaded to this great repo.