AI model photography for fashion and accessories brands
Updated · Stillhouse
The short answer
Stillhouse casts a model once and keeps them: the same face and build appear on every product, in every frame, so a lookbook reads as one campaign instead of forty strangers. You can cast an AI model or build one from photographs of a real person who has consented, and every frame is checked for both the product and the face.
The problem with AI models: a new person every time
Generic image tools invent a person on every generation. That is fine for one picture and fatal for a catalogue: the customer scrolling your collection meets a different woman on every product, and none of them is your brand's face.
Stillhouse treats the model the way a brand treats a signed face. You cast them once, approve them, and they are frozen. Every frame is checked against that approved face as well as against the product, and a frame where either drifted is retaken.
One bag, one day, one person
The six frames below were made on the studio's own image model from one newly cast model and one product photo of the handbag. Same face, same bag, a day that moves from a doorway to a dinner table. The peonies she buys at the market stay with her, which is what makes six frames a story rather than six angles.
What you can do with a cast model
- Photograph every product on the same person, in the same set or across a story
- Use an AI model, or a real person built from photographs you have the rights to use
- Reshoot an approved frame in every colourway without re-posing
- Widen a frame for a banner or story, or mend one region, without reshooting
- Turn an approved frame into a short silent clip with a slow push-in
- Frames with a generated person carry that fact in their file metadata
Questions
Can I use a real person as the model?
Yes, with their consent. Build the model from photographs of someone you have the rights to use; the studio then keeps that person consistent across every frame.
Does it work for garments as well as accessories?
The same studio and checks apply, and the product is matched against your photo. Fit and drape on a body are the hardest thing for any image model; the check catches a garment that drifted from your photo, and that take is retaken before delivery or arrives flagged with the reason.
Do I need to label images with an AI model in them?
Rules differ by marketplace and country. Every Stillhouse frame says in its file metadata that it was made with AI, and frames with a generated person also carry the keyword for it, so the file already answers when a venue asks. See how to label AI-generated product images.