Text-only hero art: Imagen 4 Ultra or Nano Banana 2.1 on Sume
Imagen 4 Ultra costs 0.075 USD, takes no references and lists 5 ratios. Nano Banana 2.1 costs 0.10 and lists 15. Which to pick for 100 hero images.
For hero art made from a prompt alone, Imagen 4 Ultra is the cheaper pick at 0.075 USD an image on Sume, but it lists only five aspect ratios, offers 1K and 2K, and takes no reference images. Nano Banana 2.1 costs 0.10 USD, lists 15 ratios including 21:9 and 8:1, offers 512 to 4K, and accepts up to 10 references.
So pick Imagen 4 Ultra when 1:1, 16:9, 9:16, 4:3 or 3:4 is enough and you will never edit a photo; pick Nano Banana 2.1 when you need a different shape, a larger size, or a reference.
Side by side
Everything below comes from the Sume image catalog's capability descriptors and base prices, read on 2026-10-10. I do not compare picture quality, because the Sume docs publish no benchmark for either model.
| Field | google/imagen-4-ultra | google/nano-banana-2.1 |
|---|---|---|
| Base price per image (USD) | 0.075 | 0.10 |
| Aspect ratios | 1:1, 16:9, 9:16, 4:3, 3:4 | auto plus 14 ratios, 21:9 to 8:1 and 1:8 |
| Resolution tiers | 1K, 2K | 512, 1K, 2K, 4K |
| input_references | none (0 to 0) | up to 10 |
| n | 1 to 4 | 1 to 4 |
| output_format | png, jpeg, webp | png, jpeg, webp |
The arithmetic for a hundred heroes
At base price, 100 images cost 100 x 0.075 = 7.50 USD on Imagen 4 Ultra and 100 x 0.10 = 10.00 USD on Nano Banana 2.1, a gap of 2.50 USD. The base price is a floor; a higher tier on a model that lists several can raise the figure, so read the endpoint pricing for your exact settings before you commit a large batch.
A gap of 2.5 cents an image is small next to the cost of redoing a batch because the frame was wrong, so let the required shape decide first.
What each model cannot do
Imagen 4 Ultra is text-to-image only in this catalog. Its input_references range is 0 to 0, so a request with a reference photo returns 400 unsupported_parameter and the message that the model is text-to-image only. It also lists no auto ratio and tops out at 2K.
Nano Banana 2.1 does not have those limits, but it costs 2.5 cents more per image at the base price and offers 512 as a cheap draft tier. If a hero image has to be reworked later from a photo or a logo, the only one of the two that can take that input is Nano Banana 2.1.
When the shape decides
A website hero is often wider than 16:9. Imagen 4 Ultra stops at 16:9, so a 21:9 banner would have to be cropped from a 16:9 image. Nano Banana 2.1 lists 21:9 natively, and also 4:1 and 8:1 for thin strips.
If you request a ratio the model does not list, Sume does not substitute a close one for you. It answers 400 invalid_request, and details.supported carries the accepted values. Reading the list first saves the failed call.
- Need 21:9, 4:1 or 8:1: Nano Banana 2.1.
- Need 4K: Nano Banana 2.1.
- Need to start from a photo or logo: Nano Banana 2.1, since Imagen 4 Ultra rejects references with
400 unsupported_parameter. - Need plain 16:9 or square art at the lowest cost of the two: Imagen 4 Ultra.
Sources
Related posts
More in Comparisons
- TikTok ad profile photo is 98x98 under 50 KB: what Sume cannot output
TikTok in-feed ads want a 98x98 px profile photo under 50 KB. No Sume route outputs that size, so generate the picture and resize it outside Sume.
- TikTok ad rejected for undisclosed AI: minor vs significant edits
TikTok's ad policy separates minor edits like lighting, background removal and denoising from significant AI changes. See where Sume's trim, crop and dim sit.
- Veo 3.1 reference images and extension: Google yes, Sume text-only
Google's Veo 3.1 takes up to 3 reference images and extends clips 7 seconds at a time. Sume's Veo rows are text-to-video only. What to use for references.
- Vidu Q4 clip lengths 3 to 16 s vs every Sume video row's limits
Vidu's image-to-video page allows 3-16 seconds. Sume has no Vidu row; this table maps that range to each Sume row's minimum and maximum, and who reaches 16.
Written by Sume