
One photo you already have.

A shot you did not take, of the same person.
A person, a product, a location — whatever must be recognisable. Several angles help. References are resized for you and cost nothing extra.
Setting, wardrobe, light, mood. The references hold identity, so your words are free to change everything around it.
Make as many variations as the campaign needs. Choose normal or high quality depending on whether you are exploring or producing a final asset.
One consistent face, product or place across a whole set
No model training and no waiting to start
Your real product, not an approximation of it
Reference images add nothing to the cost
Several references combine into one shot
Normal and high quality tiers for drafts and finals
Feeds straight into video generation as a reference
Works from ordinary phone photographs

Same face, new setting.

Same face, new wardrobe and light.

Same face, across a whole set.
One is enough to start. Several taken from different angles give the model more to work from and tend to hold identity better across a varied set of shots.
No. The references are attached to each request directly, so there is no training step and no delay before your first image.
No. References are resized before upload so a full set sits inside the free allowance — they cost nothing beyond the generation itself.
Yes. Anything that has to stay recognisable works the same way — a bottle, a garment, a room. The reference holds its appearance while the scene around it changes.
Yes. A generated image can be handed to the video generator as the opening frame or as a reference, which is how you keep one face across both stills and motion.