Nano Banana 2.1 2:3 tall image: 848x1264 to 3392x5056, then 1000x1500

Nano Banana 2.1 returns 2:3 at 848x1264 for 1K, 1696x2528 for 2K and 3392x5056 for 4K, per Google's size table. For a 1000x1500 tall pin, use the 2K frame, scale by 0.5934 and trim about 6 px of width.

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 2:3 sizes?

2:3 is the mirror of 3:2 and the sizes are mirrored too.

How do I get exactly 1000x1500?

A true 2:3 is 0.6667 and the frame is 0.6709. Scaling 1696x2528 by 0.5934 sets the height to 1500 and the width to 1006.3, so trim 6 px of width. The 1K frame would be scaled up by 1.18 times to reach 1500 tall, which is why 2K is the tier for finished pins.

If you do not want to crop at all, ask for 4:5 or 3:4 instead and pad the canvas; both are on the Nano Banana 2.1 list on Sume.

• Scale first, then center-crop the width.
• Keep titles and faces off the outer 1 percent.
• 2K is also big enough to export 750x1125 for a lighter file.

Is 2:3 on the Sume list?

Yes. Nano Banana 2.1 lists 2:3, and so do the FLUX.2 rows, Seedream and Ideogram 4.5. The GPT Image 2.x rows do not list 2:3; they list 3:4 and 4:5 for tall frames. Check the catalog row for the model you call.

{
  "model": "google/nano-banana-2.1",
  "prompt": "Autumn tablescape, overhead, warm light",
  "aspect_ratio": "2:3",
  "resolution": "2K"
}

What does it cost on Sume?

The 2K tier bills $0.15 per image, billed all-or-nothing. The image models docs show how to read it from the endpoints route.

A tall image is read on a phone, so check the crop at the size it will be shown, not at full resolution. Put the title text in the top third, and keep it out of the outer 1 percent so the 6 px trim cannot touch it. If you make several pins from one scene, change the prompt wording per pin rather than reusing one image.