12 thumbnail variants per video: cost on Sume by model

Answer

Twelve thumbnail variants per video cost $0.45 on Flux 2 Pro, $0.60 on Seedream 4.5 and $1.20 on Nano Banana 2 at 1K, using Sume's catalog per-image prices on 2026-10-04. For 30 videos a month that is $13.50, $18.00 and $36.00.

Use one frame of the video, or a headshot, as an input_references entry, and write twelve short prompts that change only the background, the expression cue or the text area. Each variant is a separate generation, billed when it completes.

Setup

• Pick a 16:9 ratio. Every row in the table lists 16:9, and Nano Banana 2 also lists 21:9 if you need an ultrawide banner.
• Send n: 4 three times per video on models whose n range tops out at 4. That is three requests instead of twelve.
• Keep text out of the render when you can. Ideogram 4.5 at $0.075 medium is the catalog's text-rendering row; the rows above are better at scenes than at long captions.
• Add the title text in your own editor or code after you pick the winner, so a typo never ships.

Cut the spend

Draft six variants on the cheapest edit row, pick two, and rerun only those two on a higher-priced model. Six drafts on Flux 2 Pro plus two finals on Nano Banana 2 costs $0.225 + $0.20 = $0.425 per video, a little less than twelve Flux 2 Pro images, with the better-looking pair at the end.

For larger batches, mode: "async" returns a job id at once and lets you poll, which avoids holding twelve connections open. See the Sume Image API docs for the request shape.

Before a large run

Prices and descriptors change when the catalog changes, so confirm them before you spend. Call GET /v1/images/models/{id}/endpoints for the row you plan to use and read its pricing line and supported_parameters; both come back in one response.

Then run a pilot of three to five images and read usage.cost on each response. Multiply by your planned count for a forecast you can trust. Completed generations are billed in full and failed or cancelled ones are not, so a pilot that errors costs nothing.

For big batches, use mode: "async" or mode: "webhook" with a public HTTPS webhook_url, so no request waits on the 30-second sync limit. Poll GET /v1/jobs/{id}/status and fetch GET /v1/jobs/{id}/result when the job completes.