YouTube swaps a 16:9 thumbnail on vertical videos: make 9:16 and 4:5

If you upload a 16:9 custom thumbnail to a vertical video, YouTube replaces it with an auto-generated 4:5 thumbnail on the home, explore and subscription pages. Your custom image still appears on the watch feed, watch history and non-mobile platforms. That is from YouTube Help (read 2026-10-09). The practical answer is to make two files per Short: a 9:16 image for the places that use it, and a 4:5 image that matches the crop YouTube would otherwise invent.

Sume's Image API can produce both ratios in one request each. This post prices that pair, using the 17 normalized ratios in the Image API docs, which include 9:16 and 4:5.

What YouTube says about each shape

For Shorts, YouTube recommends a 9:16 aspect ratio, a resolution of 2160 x 3840, and a minimum height of 640 pixels. For videos it recommends 16:9. Podcast playlists use 1:1. Accepted formats are JPG and PNG.

Two images per Short on Sume

Nano Banana 2.1 at the 1K tier is $0.10 per image on Sume (list price times 1.25). A 9:16 image plus a 4:5 image is 2 x $0.10 = $0.20 per Short. For 12 Shorts a month that is 24 images and 24 x $0.10 = $2.40. At the 2K tier ($0.15) the same plan is 24 x $0.15 = $3.60.

Set aspect_ratio explicitly. The Image API docs list 9:16 and 4:5 among the normalized values, and a model accepts only the values its catalog descriptor lists, so read supported_parameters for the model you pick. A request with a value the model does not list returns 400 unsupported_parameter, which Sume does not silently drop.

• 9:16 request: aspect_ratio 9:16 for the Shorts shelf and watch feed
• 4:5 request: aspect_ratio 4:5, which the docs note is 1080 x 1350 Instagram portrait, not 4:3
• Cost per Short at 1K: $0.20; at 2K: $0.30

Keep the face inside both crops

The two files are not interchangeable crops of one image. A 9:16 frame is much taller than 4:5, so a subject placed for 9:16 can sit lower or higher than a 4:5 frame shows. Prompt for the subject in the middle third and keep title text away from the top and bottom edges. If you prefer one prompt for both, generate the 9:16 image first and pass it as an input reference for the 4:5 version with aspect_ratio set explicitly, rather than relying on an automatic crop.

Before you publish, check which surface your viewers use. YouTube says the custom image still shows on the watch feed and on non-mobile platforms, so the 9:16 file is not wasted.

Budget for a month of Shorts

Plan the monthly number first. At 12 Shorts a month and two images each, Nano Banana 2.1 at 1K is $2.40, at 2K it is $3.60, and at the 4K tier ($0.20 per image) it is 24 x $0.20 = $4.80. YouTube's recommended Short thumbnail is 2160 x 3840, which is bigger than a 1K image, so a 1K file is acceptable only if it clears the 640-pixel minimum height and you are comfortable with the sharpness.

If you retry a failed call, Sume documents that failed or cancelled generations are not billed, so a retry costs nothing extra unless it succeeds. Keep an Idempotency-Key on async jobs so a client timeout does not create a second paid image.