A 1-step ResShift super-resolution student as a pixel-space IMAGE β†’ IMAGE upscaler for ComfyUI, at Γ—4 or Γ—2 (a scale dropdown on the loader). The student is distilled (paper 2603.22490, repo, One-Step Residual Shifting Diffusion via Distillation) from a 15-step ResShift teacher, so it does the whole super-resolution jump in a single stochastic network step β€” no multi-step sampler, no CFG.

It is model-agnostic: drop it after VAEDecode / before SaveImage and it upscales any RGB image, not just Anima / diffusion output.

… β†’ VAEDecode β†’ IMAGE ─┐
                       β”œβ”€β–Ί ResShiftUpscale ─► IMAGE (Γ—scale) β†’ SaveImage
ResShiftLoader β”€β”€β”€β”˜
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Paper for sorryhyun/Distilled-ResShift-2x