Seedream 5.0 Lite or Nano Banana 2.1 for product shots on Sume

For a plain product shot, Seedream 5.0 Lite is the cheaper pick on Sume: $0.04375 billed per image against $0.10 for Nano Banana 2.1 at 1K, which is 2.29 times as much. Choose Nano Banana 2.1 when you need what Seedream does not list: a 2K or 4K resolution tier, the 4:1, 1:4, 8:1 and 1:8 banner ratios, or aspect_ratio: auto. For 1:1 and 4:5 listing images with no special requirement, the Seedream row costs less than half.

These are Sume list facts. This post makes no claim that either model looks better on your product, and it does not rank them for quality. It tells you what each costs, what each accepts and where the difference in your bill comes from.

Price side by side

Seedream 5.0 Lite is a flat card: $0.035 list per image, times 1.25 billed. Nano Banana 2.1 is priced by resolution. The 1K row is the like-for-like comparison with a 1K product photo, and the 0.5K row is the draft tier.

What each model accepts

Seedream 5.0 Lite lists nine ratios: 1:1, 16:9, 9:16, 4:3, 3:4, 4:5, 5:4, 3:2 and 2:3. It has no resolution tier and no quality tier; for exact sizes use image_size with {width, height}. It takes up to 4 images per call and accepts reference images on edits. Nano Banana 2.1 lists 15 ratio values, including auto, 21:9 and the four extreme banners, takes the same 4 images per call, and has the four resolution tiers, 0.5K, 1K, 2K and 4K.

Both are text-to-image and edit models, so a pipeline that starts from a plain packshot and then asks for a new background uses input_references on either one.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance-seed/seedream-5-lite",
    "prompt": "A steel water bottle on a pale concrete shelf, soft side light",
    "aspect_ratio": "4:5",
    "n": 4
  }'

When the extra $0.056 per image is worth it

The gap is $0.05625 per image at 1K. Over 1,000 images that is $56.25, which is cheap if the tier you need is only on the Banana row. Three cases justify it: a 4K master you will crop for print, a banner at 4:1 or 8:1 that Seedream cannot produce natively, and an edit chain where you want auto to keep the reference's shape.

If none of those apply, spend the difference on more takes instead. At $0.04375, $10.00 buys 228 Seedream images; the same $10.00 buys exactly 100 Nano Banana 2.1 images at 1K. Two and a bit takes per Banana image is usually a better use of a product-photo budget than one pricier take.

A test that costs under a dollar

Pick ten SKUs, write one prompt per SKU, and render each on both models at 1:1. That is ten Seedream images at $0.4375 and ten Nano Banana 2.1 images at $1.00. Put the pairs in a grid and have a person who knows the product choose, then decide. If the Seedream set is good enough, the standing cost of a 500-SKU catalogue is $21.88 against $50.00.

Two things the price table hides

Draft tiers change the comparison. Nano Banana 2.1 bills the same $0.10 at 0.5K as at 1K, so there is no saving in rendering small; the only way to spend less per image is to move to Seedream or another cheaper row. If you were planning to draft on a low Banana tier and finalise at 4K, plan for $0.10 drafts and $0.20 finals.

Aspect ratio does not change the Seedream price. Its card is flat per image, so a 4:5 image costs the same $0.04375 as a 1:1. That is simpler to budget than token-billed models, where the shape changes the bill.

If you send reference photos of the product for a background change, test both models on the same five SKUs and compare edge quality and colour fidelity by eye. Price tells you the cost of a test, and it does not tell you which output your customers will prefer, so budget one small experiment before you commit a catalogue to either row. The experiment costs under a dollar at the prices in the table.