Combining multiple reference photos into one image

Sometimes everything you need is spread across different photos: the right face in one, the perfect outfit in another and the setting you want in a third. Modern AI models can combine those references into one new image. It only works well when you give each photo a clear role.

Give each reference a role

The biggest mistake is uploading several photos and hoping the model turns them into something nice. The model works better when you spell out what each reference should contribute.

• Face or person: the identity, meaning face shape, eyes, skin tone and hairline.
• Outfit or clothing: the garment, fabric and colour you want to see.
• Scene or background: the environment, location and setting.
• Light and mood: the light direction, colour temperature and overall style.

You'll usually work with two or three references. More photos don't automatically mean a better result; they actually make it harder for the model to decide what should come from where.

Tell the model what comes from where

In your prompt, state explicitly which photo provides which element. A useful pattern is: "Take the face from the first photo and have that person wear the outfit from the second photo, in the setting of the third."

Also name what has to stay stable. If identity matters, spell out the face structure, jawline, eyes, nose and skin tone. That tells the model not to reinterpret them while it adjusts the rest.

Watch out for piling on style words. Terms like "luxurious" or "hyper-detailed" can overpower the face, so the person ends up looking less like themselves. Keep the description concrete.

Make sure your source images fit together

The better your references match, the more natural the result. The model has to invent less to make the photos agree.

• Pick photos with a similar light direction. Combining a face lit from the side with a backlit scene will clash visibly.
• Mind the perspective. A face shot straight on sits badly on a body photographed at an angle.
• Watch the proportions. Very different image formats can stretch or distort parts of the result.

Also choose one reference to set the tone for light and colour, and let the others adapt to it. That keeps each element from bringing its own mood.

Build it up step by step

Want the face in a new outfit and a new setting and a different pose? Don't ask for all of it at once. The model does better when you work in steps and change one thing at a time.

1. Combine the core elements first, for example the face in the setting you want.
2. Review the result and use it as a new reference.
3. Then add the outfit or extra details.

That keeps you in control and shows you straight away which step improves things and which one ruins them. It also saves renders, and therefore cost.

How to do it in AI Formule

For combining a face with a body, outfit or scene, Swap Studio is built for exactly that. You upload a face photo and a reference for the body, outfit or setting, and choose how much of the original stays: from a full swap to just the face, or you describe your own scene in text.

If you want to work more freely with several references and your own prompt, use image-to-image in the photo generator. Upload your reference, describe what should come from where and generate a new version. You can then bring the best image to life with the video generator.

With both tools you pay per result from freely topped-up credit, with no subscription. That way you can calmly try a few variations until the combination clicks.

Ready to bring your own references together? Create an account and combine your first image.