118 likesNew · 9d old
New 8-MP text-to-image model runs in 8 steps, no CFG
What is P-H3-Image?
P-H3-Image is a diffusion model that creates photorealistic 8-megapixel pictures from a short prompt and can edit up to eight supplied images. It generates images in eight sampling steps, includes a Qwen-based prompt upsampler, and does not use classifier-free guidance.
Who it is for
Artists, developers and anyone needing high-resolution AI images can generate or modify photos without tuning CFG, as long as they have a GPU with 80-100 GB memory.