AI Image Reference: Combine Photos into One

On this page

Two references, one scene
How to use image references
FAQ

The short answer

Add each element as a separate image reference — a product, a subject, a setting — then prompt how they relate rather than re-describing them. Kling borrows the look from each and generates one new scene from all of them. It re-draws rather than composites exact pixels, so generate a few and confirm each element still reads true.

Two references, one scene

Two separate photos went in — a teapot on its own, and an empty garden table. Kling combined them into one scene, placing the teapot on the table and matching the light. All real, unedited generations:

A porcelain teapot with blue floral patterns on a plain studio background. — Reference 1 — the object

An empty rustic garden table with a lace cloth in a flower garden. — Reference 2 — the setting

The teapot from reference 1 placed on the garden table from reference 2, combined into one AI scene. — Combined into one scene

Separate shots, one believable scene. The teapot and the table were never photographed together — the AI took the look of each reference and built a single new image around them.

How to use image references

Gather your reference images

Image reference is Kling's mechanism for showing instead of describing: each picture you attach carries a subject, a setting or a style into the generation. Pick the elements you want together — clear, uncluttered references combine far more cleanly than busy ones.

Add them in the Image Reference slot

In Image Generation, the Image Reference slot takes up to ten images at once. Load each one, then describe how they relate — the AI borrows the subject and look from each and builds a single new scene from them.

Prompt the relationship, not the objects

Let the references carry what things look like; use words for how they come together — “the teapot sitting on the garden table, morning light.” Describe placement, scale and light rather than re-describing each object.

Generate a set and verify each element

Combining re-renders everything, so make several and check that each reference survived — the right product, the right setting, recognisable. Keep the take where both elements read true; discard any where one drifted.

It re-draws, it doesn't paste

Multi-reference borrows the subject and style from each image and re-renders a single new scene — it doesn't composite the exact pixels of your photos. That's what lets unrelated elements come together convincingly, but it means fine detail can shift: a label, an exact pattern, a precise face. Keep references clear and few, describe how they relate, and check every result so each element still reads as the real thing.

Frequently asked questions

What is an AI image reference?

A picture you attach to a generation so the model copies what it looks like instead of guessing from words — a product, a face, a setting, a style. Attach two and describe the scene they share, and Kling redraws one new image in which both appear, with lighting and perspective consistent rather than obviously composited. That combining move is what this page demonstrates.

How do you combine two photos using AI?

Attach both images as references, then write what the combined scene is — who is where, what the setting is, what the light is doing. Describing the relationship between the two subjects matters more than describing either one.

How do you combine photos with AI so faces still look right?

Use clear, front-lit source photos and keep the requested scene simple. Faces degrade fastest when the model is also juggling a complicated background, an unusual angle and a style change all at once.

Will the result look like a real photograph of both people?

It will look plausible, not documentary. The output is a new image, so treat it as an illustration of the two subjects together rather than evidence that they were.

AI image reference: photos, combined

Two references, one scene