A local, free pipeline that replaces slow ray-traced rendering with a single SDXL image-to-image pass. Blender exports a deliberately flat image (flat-shadowless-high-contrast-base-render). ComfyUI then adds lighting, materials and realism. The source (Stefan, 'Render anything x100 faster for free with AI | ComfyUI + Blender') claims about 10 seconds per image, or 3-4 seconds on a very good GPU, which he calls roughly 100 times faster than a ray-traced render.
The labour is split. Blender supplies structure and colour. The AI supplies light and material. Colour is deliberately carried through several channels at once (redundant-colour-carry-through-in-img2img).
The approach is not limited to interiors. A screenshot of a flower vase also works if the prompt-editing instruction is adapted to the subject. Sketch-style inputs, such as a SketchUp screenshot, are another case.
Free AI renderer ComfyUI workflow: setup and single-reference run, Generative AI rendering with control nets, reference image and prompt, Free AI preparation workflow (outline and depth from a viewport render), AI rendering trade-off: speed for control (previs use case), base-model-choice-by-addon-ecosystem.