The ComfyUI product-photo workflow chains several stages. Each stage fixes the weakness of the one before it, so the output moves from a rough scene to a faithful, relit product shot in one click.
Restoring detail before relighting would lose it again. Placing it last guarantees the final image carries the real product's detail with the new lighting.
In product photography you sell the actual product, not a 90% replica. Baioni names two obstacles to generative AI in production: incoherent lighting, and a product that is not the actual product. Generated shots a month earlier were good but unconvincing, because the lighting did not match and the subject looked copy-pasted. This chain addresses both.
The detail-restoration step came out of a community exchange. A viewer challenged Baioni on a live stream to keep a whiskey bottle's fine details. Baioni proposed Photoshop frequency separation. Reddit user Powered_JJ then posted a native ComfyUI version, and together they integrated it into Baioni's existing relight workflow.
Baioni says the demos are not cherry-picked and that the progression through the stages is organic. Limits are in ic-light-detail-restore-workflow-limits.