fusor-ai Iris-3B

Images, depth and restoration, rendered on your own CPU

Iris-3B, speridlabs' open model that draws in pixels rather than latents, compiled to WebAssembly with fusor-ai, with Qwen3-VL reading the prompt. Describe an image and watch it form in four steps with Turbo or up to a hundred with the base model, estimate a photo's depth, or enlarge it four times. Once loaded, nothing you type or choose leaves the page.

Mode