The first good and practical "pixel space" diffusion model.

[...] we propose to model the patch-wise decoding with neural field and present a single-scale, single-stage, efficient, end-to-end solution, coined as pixel neural field diffusion~(PixelNerd)

(quote from https://arxiv.org/abs/2507.23268)

https://huggingface.co/lodestones/Chroma1-Radiance
https://huggingface.co/lodestones/chroma-debug-development-only/tree/main/radiance (training checkpoints)

Updates on xitter https://xcancel.com/LodestoneRock/status/1961647453113119190

comfy pr: comfyanonymous/ComfyUI#9682

Images others have generated:

more:
lodestone-rock/flow@bd22c20#diff-112a6b6f5873ea3874f8dbc88f5ffebeb83ed7cdddc0a79df2ee8de31506b9e4

props to @lodestone-rock