How can I change the skin tone or ethnicity of my fashion models using AI?

Quick Answer

The reliable way to change a model's skin tone or ethnicity in AI product photography is to swap the on-figure person rather than try to repaint a single image. In Nightjar, pick a different Fashion Model from the 80+ pre-built roster, build a custom Fashion Model from a few reference photos, or write a short Custom Directions note like "South Asian woman, mid-20s," then rerun the shot. The garment, lighting, and framing stay more stable than a prompt-only redo.

Important caveat: changing the apparent ethnicity of an existing person in a previously-shot image can read as inauthentic representation or digital blackface, and brands have been criticised publicly for using AI as a shortcut to "diversity" instead of casting and hiring a more representative roster of real people. Use this for catalog variety with intent, not to manufacture representation you do not actually have.

Why this is a sensitive area

Representation in fashion imagery is a real commercial and ethical question. The honest baseline is to cast and pay diverse models. AI is useful when budgets, timelines, or markets make that hard for every shot, but it is not a substitute for real representation work.

Recent industry context worth knowing:

The Levi's and Lalaland.ai partnership in 2023 drew immediate backlash for using AI-generated diverse models to "supplement" their human casting.
In August 2025, an AI-generated Guess campaign in Vogue was criticised as "artificial diversity" and a way of cutting real models out of the work.

The pattern in both: audiences notice when representation is generated rather than hired, and the brand pays the trust cost. Use AI model variation to widen what your real catalog can show, not to paper over a non-diverse roster.

How Nightjar approaches this

Nightjar treats the on-figure person as a separate, reusable ingredient called a Fashion Model: an AI person you can pick, swap, or build from references and reuse across image-creation requests (Nightjar calls each one a Generation). The garment in your product photo is anchored separately, so swapping the Fashion Model leaves the product, lighting, and composition more stable than asking a prompt-only tool to redraw the whole image.

You have three paths, in order of control:

Swap to a different pre-built Fashion Model. Nightjar ships 80+ Fashion Models spanning age ranges, gender presentations, and a range of ethnicities and skin tones. Choose one that matches the audience for the shot and rerun the same Generation.
Build a custom Fashion Model from references. Upload 1 to 5 source images of the person you want, set name, age range, and gender as metadata, and Nightjar saves the Fashion Model in your Library for reuse. If you base a custom Fashion Model on a real person, only do so when you have the right to use that likeness.
Add a short Custom Directions note. Custom Directions are user-written instructions layered on top of your selected ingredients. A clause like "South Asian woman, mid-20s" or "deep brown skin tone, natural light" nudges the result without rewriting the whole brief. Broad descriptors land more reliably than precise ones.

To keep the rest of the shot aligned across SKUs, save the Photography Style (camera, lighting, and mood), Composition (framing, angle, and pose), background, and output settings as a Recipe: a saved Create-form setup you can reapply to the next product. This is how a brand can show the same garment on several Fashion Models with different ethnicities while the photographic look stays coherent across the catalog.

A practical caution on representation

Fashion Models are reusable image-generation ingredients, not a substitute for hiring real people from the communities you want to reach. If your imagery shows ethnicities and skin tones your team, casting, suppliers, and product range do not actually serve, audiences will notice. The defensible use is honest variety in service of a real catalog and a real customer base, presented as catalog options rather than as evidence of diversity work the brand has not done.