AI cold drink product photo: condensation and ice from a packshot
For a cold drink image, send the real bottle as a reference and prompt condensation, ice and rim light. Check the label is unchanged after every pass.

For an AI cold drink product photo, send your real bottle or can as the reference, prompt for condensation droplets, ice and a cool backlight, and tell the model to keep the label, cap and shape identical. Condensation hides detail, so check the label text after every pass and compare it with the source.
Prompt the effect, not the product
Describe the scene: cold surface, tiny droplets, a few ice cubes, a dark gradient, a cool rim light and a soft reflection. Keep a separate sentence that lists what must not change: the label artwork, the logo, the color and the proportions.
curl -X POST https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"google/nano-banana-2","prompt":"Keep the bottle identical: label artwork, logo, cap, shape and color. Show it cold with fine condensation droplets, a few ice cubes at the base, dark blue gradient background, cool rim light from behind, soft reflection on a wet surface.","aspect_ratio":"4:5","resolution":"2K","input_references":[{"type":"image_url","image_url":{"url":"https://example.com/bottle.png"}}]}'What condensation breaks
Droplets can bead over the label and make the model smudge or redraw the print. Check the brand name, the volume text and the small legal line. If the label changed, reduce the droplet request near the label, or generate the cold scene and composite the original label back on.
Use drafts then a final
Nano Banana 2 lists 0.5K, 1K, 2K and 4K. Try the idea at the smaller tier, choose the best composition and then regenerate at 2K or 4K. Nano Banana Pro is the heavier option when fine texture matters, and Sume's catalog prices it above Nano Banana 2 on the endpoints call.
| Model id | Aspect ratios include | Resolution tiers |
|---|---|---|
| google/nano-banana-2 | 4:5, 1:1, 16:9, 9:16, 4:1, 8:1 | 512, 1K, 2K, 4K |
| google/nano-banana-pro | 4:5, 1:1, 16:9, 9:16, 21:9 | 512, 1K, 2K, 4K |
Review checklist
- Label text matches the packshot.
- Ice and droplets look physically placed, not pasted.
- Reflection and shadow agree with the light direction.
- Cap, neck and base shapes are unchanged.
- The image is marked as a rendered scene wherever a platform asks.
Keep the edit lineage
Re-run each variation from the original bottle photo, not from the last output, so small errors do not stack. Related product notes are in AI product photo retouching.
Related posts
More in Use cases
- AI event ticket art with a working QR code: add the code after
Generate the ticket art without a code, then draw the QR or barcode on top with a library. A model-drawn code looks real but will not scan.
- AI eyewear product photos: lens glare and frame edges
Glasses are hard for image models: thin frames and clear lenses. Send a sharp packshot and ask for the lenses and edges to stay unchanged, then inspect them.
- AI backdrops for TikTok Shop: staging must match what you sell
TikTok Shop's page says LIVE backgrounds and staging items must match the products sold. How to review a generated backdrop and use video filter dim or crop.
- AI kitchen counter product photo: scale cues for the right size
Image models often guess size. Give a real dimension in the prompt and include familiar objects of known size next to the product to keep scale believable.
Written by Sume