500 SKUs with 6 images each: what a marketplace catalog costs on Sume

A 500-SKU catalog with six images per SKU is 3,000 images, which costs $22.125 on gpt-image-2.5 low at 1024, $131.25 on Seedream 5 Lite and $300.00 on Nano Banana 2.1 through Sume.

With many images per listing, the per-image price decides the plan.

3,000-image bill

Batch loop with a spend ceiling

Sume reserves the price at submit and a short balance returns 402 insufficient_credits, as described in Generation admission. Stop the loop on that status rather than retrying.

import os, requests

H = {"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"}
skus = ["tee-001", "tee-002"]
angles = ["front", "back", "detail", "flat lay", "on model", "in a box"]
for sku in skus:
    for a in angles:
        r = requests.post("https://api.sume.com/v1/images", headers=H, timeout=60, json={
            "model": "openai/gpt-image-2.5", "quality": "low",
            "image_size": {"width": 1024, "height": 1024},
            "prompt": f"Cotton t-shirt {sku}, {a} view, white background",
        })
        if r.status_code == 402:
            raise SystemExit("out of credits")
        print(sku, a, r.status_code)

Keep the work cheap

Use n (1 to 10 per call; each model has a lower ceiling, so read its n descriptor in the catalog) for variants in one call, draft at the cheapest row, and spend on a higher row only for the hero image. For video listings, stitch the stills with Timeline.

Planning notes

• Draft pass at gpt-image-2.5 low for all 500 SKUs: $3.6875.
• Final pass on the best angle only (500 images) at medium: $8.25.
• Rate and concurrency limits apply when workers pick up jobs; see Generation admission for queue capacity.
• Read GET /v1/images/models/{model_id}/endpoints for the live pricing line before a large run: prices are catalog data and can change.