Tumbler or water bottle photos with AI: composite the logo in code

A plain stainless tumbler is easy to render, and then the buyer wants their logo on it. Models redraw logos from memory, which changes a curve or a letter. A logo is an exact asset, and the right tool for exact assets is code.

The split: generate the bottle scene with Sume's Image API, then place your transparent PNG on the bottle.

Generate the blank bottle

Send a clean photo of the blank tumbler as a reference and ask for a gym bag, desk or hiking scene. Say the bottle has no printing, so there is nothing to garble.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
resp = requests.post("https://api.sume.com/v1/images", headers=H, timeout=90, json={
    "model": "openai/gpt-image-2.5",
    "prompt": "Place this exact blank stainless tumbler on a desk beside a laptop, soft light. The tumbler has no printing or logo.",
    "input_references": [
        {"type": "image_url", "image_url": {"url": "https://example.com/tumbler-blank.jpg"}}
    ],
    "aspect_ratio": "auto",
})
resp.raise_for_status()
if resp.status_code == 202:
    raise SystemExit("job envelope: poll data.status_url")
body = resp.json()
print([d["url"] for d in body["data"]], body.get("usage"))

Place the logo

Measure the logo area on the saved image, fit the logo to it and paste with the PNG's alpha channel. A slight reduction in opacity helps it sit on shiny metal.

from PIL import Image

scene = Image.open("tumbler-scene.png").convert("RGBA")
logo = Image.open("logo.png").convert("RGBA")

w = 180
logo = logo.resize((w, int(logo.height * w / logo.width)), Image.LANCZOS)
alpha = logo.getchannel("A").point(lambda a: int(a * 0.92))
logo.putalpha(alpha)
scene.alpha_composite(logo, (455, 610))
scene.convert("RGB").save("tumbler-final.jpg", quality=92)
print("ok")

When the bottle curves

A flat paste suits the front of a straight tumbler. For a wide logo on a rounded body, squeeze the logo horizontally toward the edges, or generate with the logo side turned three quarters to the camera and accept a smaller mark. Check the result against the logo file at 100 percent, and keep usage.cost from the image call with the final asset.

Before you run a catalog

Run one SKU end to end first. POST /v1/images returns the images directly when it finishes within 30 seconds; past that it returns a 202 envelope with status_url and result_url, which the jobs and results guide explains. Handle that branch before you loop over a catalog, and write each result's URL and usage.cost to a file keyed by SKU, so a failed run restarts where it stopped and nothing is paid for twice.