A free MOSS-Image adapter places people and objects from a flat colour map and a depth map
MOSS-Image Color-Depth Control is a free community adapter that makes the open MOSS-Image model follow a colour map of where each object sits and a depth map of how far away it is. You can block out a scene with 3D mannequins and get a finished picture with the same placement, scale and pose, while the look and background come from your prompt.
What it does
The colour map gives each object or region its own flat colour (the colours are only IDs, black means unspecified) and the depth map is greyscale with near as white; either one can be left out. Generation still starts from pure noise, so unlike image-to-image from a 3D render it does not carry over the render's grey surfaces and flat background. In ComfyUI the adapter loads with the normal LoRA loader and the MOSS-Image Color/Depth Control node sits between the text encode and the sampler, with strength, start and end percent.
Results and limits
The author's grids show the trade-off openly: with the control on for the whole run, the mannequin's surface shows through, while ending it at half the steps puts real people where the maps say. Without per-region prompts a mannequin-shaped map gets drawn as a mannequin, and with a manga-style map the panel layout drifts and text degrades at later checkpoints. Paired with the same author's 4-step adapter, the card reports about 1.8 seconds per 1024×1024 image on an RTX PRO 6000.
Worth knowing
Recommended settings are strength 1.0, an end of 0.5 for 3D mannequin or other synthetic maps and 1.0 for maps traced from real photos, with one region prompt per colour. Four checkpoints ship (2,500 to 10,000 steps; earlier ones follow the map more loosely). Native ComfyUI support is in the v0.43.0 release. It is unofficial and Apache 2.0, not an OpenMOSS release.