AI combine two images: how to merge two photos step by step in 2026

Short answer: to use AI to combine two images, upload both photos to a prompt-based editor such as Google Gemini, tell it which photo is the subject and which is the background, and describe the final scene in one sentence. That free route handles most jobs in under a minute. If the result has to survive close inspection, for a product shot or client work, use Photoshop's Harmonize tool instead, which keeps your original pixels and matches the light for you.

This guide is the practical side of the topic. You get four step-by-step methods, prompt templates you can copy, fixes for the most common problems, and a short tool list. If you want a full ranked comparison of every app, read our roundup of the best image combiner AI tools; this page focuses on how to actually do the merge.

We put the guide together from official product pages, published tutorials and pricing pages, checked in late October 2026. Nobody at The AI Future ran hands-on tests for it, so the steps follow the vendors' own instructions and independent guides, which we link throughout.

Key takeaways

• Start with a prompt-based tool. Gemini and ChatGPT both let you merge two photos by describing the result in plain English, and both work on free accounts.
• Name each image's job. Saying "the person from the first image, the beach from the second" is the single biggest fix for muddled merges.
• Use layers when realism matters. Photoshop's Harmonize matches color, lighting, shadows and reflections while your originals stay intact.
• Build complex scenes in passes. Merge two images first, then feed the result back in and add the next element.
• Only combine photos of people with their consent. Putting someone into a scene they were never in can mislead and harm.

Which method should you use?

There are two basic ways an AI photo merge works. Prompt-based models redraw the whole picture so the two sources look like one shot. Layer-based editors keep your photos as they are and use AI only to cut out, relight and blend. Pick the method by the job, not by the tool's hype.

Method	Best for	Skill needed	Cheapest way in
Prompt merge in Gemini	Fast, free blends of people, pets, products and places	None	Free
Chat-based merge in ChatGPT	Merging, then refining by conversation	None	Free (limited)
Layer composite in Photoshop	Realistic product, real estate and client composites	Moderate	$19.99/mo
Canvas blend in a free web editor	Quick two-photo blends with some manual control	Low	Free daily credits

Choose Gemini if you want the quickest free result. Choose ChatGPT if you expect to fix the image three or four times. Choose Photoshop if the merged photo will be printed, sold or shown to a client. The rest of this guide walks through each one.

Before you start: prepare your two images

Most bad merges start with bad inputs. Five minutes of prep saves a lot of regenerating.

• Use clean, sharp sources. Runware's Nano Banana 2 composition guide notes that "a subject shot on a plain background composites more predictably than one already embedded in a busy scene." If you can, shoot the subject against a plain wall.
• Match the angle roughly. A person photographed from above will look odd dropped into a scene shot at eye level. AI can fix light more easily than perspective.
• Keep resolution similar. A tiny, blurry subject on a sharp 4K background is hard to hide, even with AI.
• Decide the roles. Before you type anything, decide which image is the subject (what goes in) and which is the background or style (where it goes, or how it looks).
• Check you have the rights. Use your own photos, licensed stock, or images of people who agreed to it.

Method 1: how to combine two images with AI in Gemini (free)

Gemini is Google's assistant, and its Nano Banana image models handle editing and merging. Google's image generation page says Nano Banana is "available for free," with paid Google AI plans unlocking the newer Nano Banana 2.1 model. That makes it the best starting point for most people who want an AI image combiner free of charge.

1. Open the image tool. In the Gemini app, Google says to open the tools menu and select "Create images", then choose Fast, Thinking or Pro from the model menu.
2. Upload both photos. Add your subject photo first and your background photo second, so the order matches how you will refer to them.
3. Write a role-based prompt. For example: "Place the woman from the first image on the beach in the second image. Keep her face, hair and clothes exactly the same. Match the warm sunset light and add a soft shadow on the sand."
4. Refine with small follow-ups. Google says you can tell Gemini to "change the background, replace an object or add an element" without starting from scratch. Ask for one change at a time, such as "make her slightly smaller" or "move her to the left third of the frame."
5. Upgrade a near-miss. Google AI Pro, Plus and Ultra subscribers can pick "Redo with Pro" from the three-dot menu to regenerate with Nano Banana Pro.
6. Download and check the watermark. Google says Gemini images carry an invisible SynthID watermark and a visible watermark showing they are AI-generated, so plan around that if you need a clean commercial file.

Google's own prompt advice is to build on a simple formula: action, subject, setting and atmosphere, then add composition, style and aspect ratio. For a merge, that becomes: what to place, from which image, into which image, in what light. Our guide to Google Gemini AI photo prompts has more templates.

Cost: free to start. As of 16 October 2026, Google listed Google AI Plus at $4.99 a month and Google AI Pro at $19.99 a month in the US, according to ProPakistani, which also reports that Plus may lose manual Pro-model selection on 13 November 2026. Check the official plans page before you pay.

Verdict: if you only learn one method, learn this one.

Method 2: merge two photos in ChatGPT and refine by chat

ChatGPT is the better pick when you know the first result will need several rounds of tweaks. OpenAI released ChatGPT Images 2.5 on 8 September 2026 and says it is "better at preserving the subjects in your reference photos" and "follows editing instructions more reliably across multiple turns." It is available on every ChatGPT tier, including free.

1. Start a new chat and attach both images. Use the upload button, then say which is which in your first line.
2. Describe the merged scene. Try: "Combine these two photos into one frame. Put the dog from image 1 on the sofa in image 2, sitting on the left cushion, lit by the window."
3. Point at problems. OpenAI says you can now place comments directly on images for focused edits. Click the spot that looks wrong, such as a floating paw or a missing shadow, and describe the fix.
4. Sketch the layout if words fail. Type "@Sketch" to draw a rough layout, which OpenAI says ChatGPT uses as a visual guide for the final image.
5. Stop while it still looks right. Even with better multi-turn consistency, long edit chains can drift. Save good versions as you go.

Cost: the free plan caps image generation. Krater's ChatGPT pricing breakdown, based on OpenAI's pricing page in October 2026, lists Go at $5.50 a month (with ads), Plus at $20 and Pro from $100. Prices vary by country.

Verdict: pick ChatGPT if you already pay for it and prefer to edit by talking rather than by clicking.

Method 3: build a realistic composite in Photoshop with Harmonize

Prompt-based tools redraw everything, which can subtly change a face, a logo or a product label. Photoshop works the other way. You place your subject on the background yourself, and the Harmonize tool uses Adobe's Firefly models to adjust color, lighting, shadows and visual tones so it fits, as TechCrunch reported at launch.

The steps below follow the Photoshop Essentials tutorial, which says you need Photoshop 2026 (version 27.0 or later):

1. Open the background image. This is the scene your subject will join.
2. Place the second image. Choose File > Place Embedded and select your subject photo. It lands on its own layer as a smart object.
3. Remove the subject's background. Click Remove Background in the Contextual Task Bar. Photoshop hides the background with a layer mask rather than deleting it.
4. Resize and position first. Use Edit > Free Transform. The tutorial stresses doing this before Harmonize, because Harmonize blends the element "at that exact size and location."
5. Click Harmonize. Photoshop generates a new Harmonize layer that matches color, light, shadows and contrast, and can even add reflections.
6. Pick a variation. Harmonize gives you three options in the Properties panel, and you can click Generate for three more.
7. Fix flat highlights by hand. The tutorial's author found Harmonize "can sometimes flatten the highlights and make the subject too dark," and shows how to paint back brightness with a masked copy of the original layer and a Color and Vibrance adjustment.

That last step is the point of this method: because every layer is still there, you can fix one detail without regenerating the whole picture. For more on the wider toolset, see our guide on how to use Photoshop AI.

Cost: no free plan. The Photography plan, which includes desktop Photoshop and Lightroom, is $19.99 a month and Photoshop alone is $22.99, according to Dupple's October 2026 price check of Adobe's US list prices.

Verdict: worth the learning curve when the composite has to pass for a real photo.

Method 4: use a free canvas editor such as Dreamina

Web editors with an AI canvas sit between the two approaches. You get a simple blend button, plus manual tools for when the AI misjudges the layout. CapCut's Dreamina is a typical example, and its own merge guide lays out the process:

1. Upload both pictures in the Canvas tab under "Generate images".
2. Select both and choose Blend, marking one image as the background and the other as the foreground.
3. Add a short style prompt, such as "warm, sunset, soft lighting, 4K," then click Generate.
4. Fall back to manual tools if needed. Dreamina suggests Remove Background, Remove, Inpaint and Expand to clean up edges or extend the frame.
5. Export as JPEG or PNG at the size you need.

Dreamina's guide says it gives 150 free credits a day and that its AI feature costs three credits per use. That is a vendor claim from an undated page, so check your account's balance before you rely on it.

Verdict: a good free option if you want a blend button and a bit of manual control without learning Photoshop.

Prompts that work for combining two images

Whatever tool you use, the prompt does the directing. The images supply the what; your words supply the where and how. Runware's guide puts it plainly: state "placement, scale, and contact" because "the references can't tell the model how the pieces relate." Copy and adapt these:

• Person into a place: "Place the person from image 1 standing on the left side of the street in image 2. Keep their face and clothes unchanged. Match the overcast light and add a soft shadow."
• Two people in one frame: "Combine the two people from image 1 and image 2 into one photo, standing side by side in a park at golden hour. Keep both faces exactly as they are."
• Product into a lifestyle scene: "Put the watch from image 1 on the wooden table in image 2, resting near the coffee cup, lit by warm morning window light."
• Style blend: "Redraw image 1 in the colors, brushwork and texture of the painting in image 2. Keep the layout and buildings of image 1."
• Pet into a portrait: "Add the dog from image 2 sitting at the feet of the person in image 1, scaled naturally, same lighting direction."

Three habits help every time. Refer to images by position and description ("the watch from the first image"). Describe the target lighting once and let the model relight both sources. Keep identity instructions explicit ("keep the face unchanged"), since that is what drifts most. For more general prompt craft, see our AI photo prompt guide.

Fixing common problems

The subject looks pasted on

This is almost always light and shadow. In a prompt tool, name the light source and ask for a contact shadow ("add a soft shadow under her feet on the sand"). In Photoshop, run Harmonize again and try the other variations. Our guide on how to make an AI photo look real covers more fixes.

The face or product changed

Prompt-based models redraw the whole image, so details can shift. Add "keep the face, hair and clothing exactly the same" or "keep the label text unchanged." If it still drifts, switch to a layer-based method, where your original pixels stay untouched.

The AI mixed up which image is which

Upload in a fixed order and refer to "image 1" and "image 2" every time. Describe each element too ("the man in the blue jacket from image 1").

The scale is wrong

Give the model a reference point: "about the height of the doorway" or "small enough to fit on the table." In Photoshop, set the size before you run Harmonize.

You need more than two images

Some platforms accept many references at once; Runware's API docs say one Nano Banana 2 call takes up to 14. Even so, the same guide recommends building busy scenes in passes: get two or three elements right, then feed that result back in and add the next one.

A short list of tools to combine two images

These are the tools this guide covers, plus the job each one suits. For full pros, cons and rankings, see the image combiner AI roundup, and for creating images from scratch, our list of the best AI image generators.

Tool	Type	Best for	Free option	Paid from
Google Gemini	Prompt-based	Free all-round merges	Yes	$4.99/mo
ChatGPT	Prompt-based	Refining over many edits	Yes (limited)	$5.50/mo
Adobe Photoshop	Layer-based	Photo-real composites	No	$19.99/mo
Dreamina	Canvas editor	Quick free blends	Yes (daily credits)	Check site

Prices are US list prices checked in October 2026 from the sources linked above, and they change with region and promotions.

Free vs paid: what you actually get

You can do a convincing AI photo merge without paying anything. Gemini's free tier includes Nano Banana, ChatGPT's free tier includes Images 2.5 with caps, and Dreamina advertises daily free credits. For social posts, invitations and fun edits, free is enough.

Paying buys you three things. First, more attempts: Google's paid plans offer higher usage limits than free, and ChatGPT's paid tiers lift its image caps. Second, better models, such as Nano Banana 2.1 on Google's paid plans. Third, control: Photoshop's layers let you fix one detail without regenerating everything. If you merge images for work every week, the $19.99 to $20 tier of any of these is where the time savings start to pay off.

Use photos of people responsibly

Combining your own photos, licensed stock, or pictures of friends and family who said yes is fine. Putting a real person into a scene they were never in without their permission is different. It can mislead people and cause real harm, and making intimate or deceptive images of real people without consent is illegal in many places. Both Google and OpenAI add watermarks or C2PA metadata to AI images, and OpenAI says it checks prompts and images to help prevent harmful outputs. If you plan to sell merged images, read whether you can sell AI-generated art first.

Final verdict

Most people should start with Gemini: upload two photos, name each one's role, describe the light, and refine with one small change at a time. Move to ChatGPT if you like to edit by conversation, and to Photoshop with Harmonize when the composite has to look real up close. Whichever tool you choose, clean inputs and a clear prompt matter more than the price tag.

FAQ

Can AI combine two images for free?

Yes. Google says Nano Banana in the Gemini app is available for free, ChatGPT's free plan includes image editing with limits, and Dreamina's guide advertises 150 free daily credits. Expect caps and watermarks on free tiers.

What is the best AI to combine two images?

For most people, Gemini, because it is free and handles subject-into-scene merges from a plain prompt. For realistic commercial composites, Photoshop's Harmonize gives more control. Our image combiner roundup ranks eight options.

How do I combine two photos in one frame online with AI?

Upload both photos to an AI editor, say which is the subject and which is the background, and describe placement, scale and lighting in one or two sentences. Then refine with short follow-up edits.

Why does my merged image look fake?

Usually because the light direction, color temperature or shadows don't match. Name the light source in your prompt, ask for a contact shadow, or use a relighting tool such as Harmonize.

Can I blend two images into one style with AI?

Yes. Use one image for content and the other for style, and tell the model which is which. Google says Nano Banana can take "the texture, colour or mood from any reference image and map it directly to your subject."

Will the AI keep faces the same?

Newer models are better at it, and OpenAI says Images 2.5 keeps reference subjects more recognizable. It is still not guaranteed, so say "keep the face unchanged" and check closely. For exact likeness, use a layer-based editor.