Turn One Flat Product Photo into Lifestyle Scenes with Image-to-Image Prompts

A white-background packshot is useful for a product page. It is rarely enough for ads, social posts, or landing pages. You usually need the same product sitting on a kitchen counter, held in a hand, or lit for evening mood — without booking a new shoot every time.

Image-to-image workflows help here: you keep the product from a reference photo, then rewrite the scene with text. Below is a practical prompt workflow you can try in any text-and-image AI studio that supports reference uploads. One option is Nano Banana CN, a Chinese-friendly AI image and video studio. It is an independent site and is not affiliated with Google.

What you need before you open the generator

1. One clean product photo. Prefer a sharp, front-facing shot on a plain background. Soft shadows are fine; heavy props in the frame make later edits harder.
2. A short brief for the new scene. Write three lines only:
   • Where the product sits (marble counter, wooden desk, outdoor table)
   • Who interacts with it, if anyone (hands only, no face, or no person)
   • Mood and light (soft morning window light, warm evening lamp, bright studio softbox)
3. Output use. Cover, square social post, or wide banner — pick the aspect ratio before you generate so the composition does not fight the crop later.

You do not need a design background. Most of the work is clarifying the brief, then iterating prompts.

Pick a model for the job (not for the brand name)

On Nano Banana CN, the image models are positioned like this:

• Nano Banana 2 — a sensible default for everyday product and lifestyle frames. Nano Banana CN lists support up to 4K for it.
• Nano Banana Pro — when the frame must carry readable labels, small type, or dense detail.
• Nano Banana 2.1 — when subject consistency matters across several references. Nano Banana CN lists better layout and subject consistency and up to 14 reference images.

If you only have one packshot, start with Nano Banana 2. Move to Pro when text on packaging must stay legible. Try 2.1 when you upload several angles of the same SKU and need them to read as one product.

Style presets such as photography, product, illustration, 3D, or pixel art help after the scene is correct. Lock product and lighting first, then switch style.

A reusable prompt skeleton

Treat the prompt as five blocks. Keep them in this order so the model does not invent a new product:

1. Task — what to change
2. Keep — what must stay from the reference
3. Scene — place, props, people
4. Camera & light — lens feel, time of day, shadows
5. Constraints — what to avoid

You can try this skeleton (replace the bracketed parts):

Edit the uploaded product photo into a lifestyle scene.
Keep the product shape, label layout, colors, and proportions from the reference.
Place the product on [surface] in [location]. Add [1–2 small props].
[No person / show only hands / show a person from behind].
Lighting: [soft window light / warm lamp / bright softbox], realistic shadows under the product.
Camera: eye-level, mild depth of field, natural photo look.
Do not change the logo text, do not restyle the packaging, do not add extra logos.

Example fill for a ceramic mug:

Edit the uploaded product photo into a lifestyle scene.
Keep the mug shape, handle, glaze color, and printed mark from the reference.
Place the mug on a light oak desk beside an open notebook and a soft linen napkin.
Show only a hand holding the mug by the handle.
Lighting: soft morning window light from the left, gentle shadow under the mug.
Camera: eye-level, mild depth of field, natural photo look.
Do not change the logo text, do not restyle the packaging, do not add extra logos.

The “Keep” and “Do not change” lines matter more than poetic adjectives. They are there to stop the model from inventing a prettier box and breaking the SKU.

Image-to-image habits to try

Studios that support reference images often let you upload a packshot and describe the new scene while preserving composition and lighting cues from the source. Practical habits:

• Upload the packshot as the main reference, not as a tiny thumbnail buried under other images.
• Say what to preserve before you describe the new room. Product identity first, mood second.
• Change one variable per retry. If the background is wrong, keep the lighting line fixed. If the logo warps, shorten the scene description and strengthen the keep/constraint lines.
• Generate a small batch when the tool allows variants, then pick one direction instead of rewriting the whole prompt each time. Changing color, background, or style on a good base frame keeps the product you already approved.
• Use templates or prompt-library starters for posters, covers, and e-commerce layouts when you need a finished marketing crop, then swap in your product-specific keep/constraint lines.

For multi-angle consistency (front, three-quarter, detail), Nano Banana 2.1’s multiple-reference path is the one to try: upload several real photos of the same product and ask for the same scene across those identities.

A second workflow: white background → scene → social crop

A common marketing path looks like this:

1. Start from a catalog cutout or plain-background shot.
2. Run image-to-image with the skeleton above to place the product in a believable room.
3. Duplicate the winning frame into two crops: a 1:1 feed post and a wider cover.
4. If the tool offers AI upscaling, run it on the final pick to improve resolution and clarity and to reduce blur, noise, or compression artifacts before you publish or print.

You can try a short “scene only” follow-up prompt after the first good result:

Keep this exact product and label. Same camera height.
Replace only the background with a [sunset balcony / marble kitchen island].
Keep shadow direction consistent with the new light.

A second pass keeps each change small and easier to check.

Prompt checks before you hit generate

Thirty-second checklist:

• Name the product category once (mug, bottle, bag).
• Include one keep sentence about shape, label, and color.
• Include one do-not sentence about logo and packaging.
• Limit people to what you need (none / hands / back view).
• Match aspect ratio to the channel.

If any answer is no, fix the prompt before another generation.

What this workflow is not

• It is not a substitute for licensed product photography when legal or brand guidelines require a studio shoot.
• It is not a promise about a specific result. The prompts above are starting points, and results vary by model and source image.
• It is not a way around image rights. Use your own product photos or assets you have rights to use.

Closing note

Start with one SKU, one scene, and the five-block prompt. When product identity holds, explore style presets and upscaling. Order matters: identity, scene, style, polish.

Disclosure: this post was published by the Nano Banana CN team and drafted with AI assistance. The example prompts are suggestions and were not run for this article.