Nano Banana 2.1 4:5 portrait: 928x1152 to 3712x4608, then 1080x1350
The 4:5 sizes Nano Banana 2.1 returns per tier and the 7 px crop that makes the 2K frame an exact 1080x1350 Instagram portrait.

Ask Nano Banana 2.1 for 4:5 and Google's table gives 928x1152 at 1K, 1856x2304 at 2K and 3712x4608 at 4K. Instagram's portrait size is 1080x1350, so scale the 2K frame down and trim about 7 px of width: the model's frame is a hair wider than a true 4:5.
Sume forwards your ratio and tier to the provider and does not document the exact pixels it returns, so this post uses Google's table. Check width and height on the first file you get back before you build a crop on them.
What are the 4:5 sizes by tier?
These come from the Google output table for this model.
| Tier | Pixels | Width divided by height |
|---|---|---|
| 1K | 928x1152 | 0.8056 |
| 2K | 1856x2304 | 0.8056 |
| 4K | 3712x4608 | 0.8056 |
Why is a crop needed at all?
A true 4:5 is 0.8, and the frames are 0.8056 because the sizes are rounded to the model's grid. Scale 1856x2304 by 0.5859 so the height is exactly 1350 and the width becomes 1087.5. Trim 7.5 px, about 4 on each side, and you have 1080x1350.
That margin is under 1 percent of the width, but keep logos and captions off the outer edge anyway.
- 1K at 928x1152 needs a 1.17 times upscale, so keep it for drafts.
- 2K is the practical tier for feed images.
target_pixelsis never applied on Sume, so scale and crop in your own code.
How do I request it on Sume?
4:5 is in the Nano Banana 2.1 ratio list on Sume, next to 5:4, 1:1, 3:4, 2:3, 9:16, 16:9, 3:2, 4:3, 21:9, 4:1, 1:4, 8:1, 1:8 and auto. Read GET /v1/images/models to confirm the row before you hard-code anything.
{
"model": "google/nano-banana-2.1",
"prompt": "Flat-lay of a ceramic mug on linen, soft window light",
"aspect_ratio": "4:5",
"resolution": "2K"
}What does a 2K image cost?
On Sume the 2K tier bills $0.15 per image, billed all-or-nothing. A 1K draft is $0.10, so a draft-then-final loop for one post costs $0.25 before retries. The image models docs list the other rows if you want a cheaper draft model.
Do the resize with a resampler that handles downscaling well, such as Lanczos, and export once. Resizing twice, for example 2K to 1500 tall and then to 1350, softens fine text. If a post needs several crops, a 4:5 feed image and a 9:16 story from the same scene, generate the story separately instead of cropping the portrait.
Sources
Related posts
More in Models
- Nano Banana 2.1 at 8:1 and 4K: 12288x1536, 18.87 megapixels
Google's size table puts Nano Banana 2.1 8:1 at 4K at 12288x1536, about 18.87 MP, larger than a 4096x4096 square. Sume lists 8:1 and 4K at $0.20 flat.
- Nano Banana 2.1 9:16: 768x1376, 1536x2752 or 3072x5504 for stories
The three 9:16 sizes Nano Banana 2.1 outputs and how to reach 1080x1920 for Stories and Reels covers with a 15 px vertical trim.
- nano-banana-2 still works but runs as 2.1: check the stored job id
Sume retired Nano Banana 2: google/nano-banana-2 runs as Nano Banana 2.1 and the job stores the 2.1 id. A Python check and what you pay.
- Nano Banana 2 to 2.1 on Google: 4K falls from $0.151 to $0.113
Google's price per image falls about 50% at 1K and 2K and 25% at 4K from Nano Banana 2 to 2.1. Sume runs nano-banana-2 requests as 2.1 at its own 2.1 rates.
Written by Sume