Pixel-space cascaded text-to-image diffusion model with a T5-XXL text encoder. Stage I (64px) → Stage II (256px) → Stage III x4 upscale (1024px). [model] [license]