Lore

Multi-mannequin composition

# Multi-mannequin composition

Copy the 3D node several times to place multiple products, or different products, in one layout reference. This controls how they touch, how close they sit and which one overlaps. These are the spatial relationships that are hardest to describe in text. The idea is Split placement from appearance: a stand-in sets geometry, the image model sets looks applied to a group.

Procedure

  1. Duplicate the 3D node, for example three toothpaste tubes, or a second bottle model alongside the first.
  2. Position and rotate each mannequin, see Garbage 3D product mannequin.
  3. Merge the posed models in a compositor into a single image A.
  4. Run the annotation step (Compositor annotation layers (labelled references)).
  5. The rest of the workflow, including Master prompt with reference-image roles (fixed scaffold), stays the same.

Takeaway

Placement replaces prose for spatial relationships. A composition that would need a long, expert text description is set up by moving objects around, and the result can feed banners or short animated reels.

Related

Scaling to several products

Scaling to several products

The downstream workflow does not change with the number of products, because the compositor merges all mannequins into one layout image. The same pipeline handled one product, three, and two different products.

Final demo. A second, differently posed bottle was added. The prompt asked for the bottles to be leaning on each other, and the result showed the bottles touching, as requested. The author was unsure of the exact phrasing that produced it. Contact between products is therefore requested in the scene text, while the mannequins set where each bottle sits and how it is turned (see Split placement from appearance: a stand-in sets geometry, the image model sets looks).