Nano Banana 2.1 16:9 output size: 1376x768 up to 5504x3072

Google's table puts a 16:9 Nano Banana 2.1 image at 1376x768 for 1K, 2752x1536 for 2K and 5504x3072 for 4K. None of those is 1920x1080, so for a 1080p slot you generate at 2K, scale down and trim 15 px. On Sume you request aspect_ratio: "16:9" plus a resolution tier.

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 exact 16:9 sizes?

Google publishes a pixel table per ratio and tier for this model. The 16:9 row is below, next to Google's own token price for the same tier.

How do I get exactly 1920x1080?

Take the 2K frame. Scaling 2752x1536 so the height is 1080 gives a width of 1935, which is 15 px too wide, so you trim about 7 px from each side. The 1K frame would need a 1.4 times upscale, so start from 2K whenever the image has fine detail such as text or product edges.

Sume does not do this step for you. Output size follows the tier and ratio you request, and target_pixels is never applied, so the resize lives in your code or your editor.

• 2752x1536 scaled by 0.7031 is 1935x1080.
• Center-crop 15 px of width to reach 1920x1080.
• 5504x3072 halved is 2752x1536, so a 4K frame is only worth it if you crop in.

What does the call look like on Sume?

Send the org/slug id. The catalog row is the source for which tiers your account sees, so read GET /v1/images/models first and pin only values it lists; anything else returns 400 unsupported_parameter.

{
  "model": "google/nano-banana-2.1",
  "prompt": "A harbor at dawn, wide shot",
  "aspect_ratio": "16:9",
  "resolution": "2K"
}

What does it cost on Sume?

Sume bills the provider list price times 1.25. Nano Banana 2.1 is $0.10 at 1K, $0.15 at 2K and $0.20 at 4K per image, billed all-or-nothing. The Google prices in the table are Google's rate card, not what Sume charges. Parameters and rows are in the image models docs.

Before you build a pipeline on these numbers, generate one test image per tier and record the width and height you actually receive. Keep that measurement in your config next to the crop offsets, so a change on the provider side shows up as a failed assertion in your code instead of as a soft-edged banner in production.