Structure and easier ussage

[IGN-Styly]

Having someone host this is clearly not viable as so, I also feel that having users having to self host is not good. As such i recommend using wgpu on web to compute locally while having an easy setup.
Go to website
Webgpu computes

[MarcelLieb]

+1 for inference setup with maybe ONNX Runtime Web.
On second thought the VRAM and compute requirements may be too high to be realistically possible.

[M1noa]

+1 this is semi-common and is more accessible to people who don't even have gpu's it just will be slow as you'll have to download and run the modal all from a browser