How can I use AI to create a consistent 'lookbook' style for a new collection?

Quick Answer

Lookbook consistency does not come from prompts. Pick three to five reference images that describe the collection's visual language, then save the entire setup (style, framing, model, output settings) as a reusable preset and apply it to every product in the collection. In Nightjar, that preset is called a Recipe, and it is what holds lighting, color, and camera feel across the set.

Why prompts alone drift

Typing "cinematic lighting" twice yields two different results. Text prompts are weak as a production system. They get interpreted slightly differently each time you generate an image, which is the source of the lookbook drift you are trying to eliminate.

What to lock down before you generate the first image

A lookbook is a system, not a single hero image. Four variables drift between AI shots and need to be locked, not re-typed:

1. Photographic look (lighting direction, color grade, lens feel, mood). Nightjar calls this a Photography Style: a reusable visual direction extracted from one to five reference images. For a lookbook, build a custom one from your collection's mood-board references rather than picking from the 150+ that ship with Nightjar.
2. Pose and framing. Nightjar calls this a Composition: a reusable arrangement that controls camera angle, crop, product placement, and the model's pose. Use the same Composition across products so the catalog grid reads as one shoot.
3. Model identity. Nightjar calls this a Fashion Model: a reusable AI person you can pick from 80+ pre-built options or build from one to five source images. Reusing the same model across the collection is what stops the person from changing between shots.
4. Output settings: aspect ratio, resolution, output format, image count.

Save it once as a Recipe, apply it to every product

A Recipe in Nightjar is a saved Create-form setup: it stores the Photography Style, Composition, Fashion Model, background, custom directions, and output settings, but not the product images themselves. This is the scale layer that turns "I made one good image" into "the next eighty products look like the same shoot."

The flow:

1. Curate three to five reference images that describe the collection's visual language.
2. Save them as a custom Photography Style in your Library.
3. Pick or build a Composition for framing and a Fashion Model when the collection is model-led.
4. Combine all of it into a Recipe with the aspect ratio, resolution, and output format you need.
5. Apply the Recipe to each product in the collection. Swap the product image, leave everything else.

What this preserves

• Lighting direction and quality across products.
• Color temperature and grade.
• Camera language and lens feel.
• Framing and pose discipline through the saved Composition.
• Model identity through the saved Fashion Model.

The collection reads as one shoot rather than a stitched set of unrelated AI generations, and the Recipe is reusable for next season's drop.