What this does
One clean phone photo in, a set of consistently lit product images out. The graph holds the product’s shape with a structure pass and carries its colour and texture with an image-prompt adapter, so the model changes the scene around the product without redesigning the product.
That separation is the whole trick. A plain text-to-image prompt will happily return a beautiful bottle that is not your bottle. Constraining structure and appearance as two separate inputs is what keeps the thing you are selling recognisable.
What you get
- The ComfyUI workflow as a
.jsonyou can drag straight onto the canvas - Model and node requirements listed up front, with what is optional and what is not
- A settings sheet: which sampler, how many steps, and what the CFG value actually does here
- Three prompt sets — warm studio, cool studio, plain seamless background
- The mask recipe for keeping a clean edge on reflective products
What it replaces
Not a photographer. It replaces the session you were going to book for a product that will look slightly different in three months, and the per-image fee you pay to get a consistent background across a catalogue.
Honest limits
It cannot invent detail that was never in the source photo. Feed it a dark, soft shot and you get a sharp dark soft shot. Photographed cleanly on a plain surface this works well; photographed on a cluttered desk you will spend longer masking than shooting.
Fine text and thin reflective edges are where it fails. Logos with small type will drift. Anything with a mirror finish needs its mask checked by eye before you ship it.