18 hero images: cutout costs $0.405, upscale costs $3.60 on Sume
Sume's image upscale is $0.20 per image and RMBG is $0.0225, an 8.9x gap. For 18 hero images that is $3.60 versus $0.405, and the order of the steps matters.

On Sume, image upscale costs $0.20 per image and background removal costs $0.0225 per image, so an upscale is 8.89 times a cutout. For 18 hero images, cutouts cost 18 x $0.0225 = $0.405 and upscales cost 18 x $0.20 = $3.60. Doing both is $4.005, or $0.2225 per image. Both are flat prices on the Sume price list (read 2026-10-08), and both are paid MCP tools named rmbg_create and image_upscale_create.
Because the upscale is almost nine times the cutout, the question is which step you can skip, not which to do first.
Prices side by side
These are the flat utility prices from the lead-verified Sume price list, next to the cost of generating a new image at a similar finish.
| Step | Price per image | 18 images | Notes |
|---|---|---|---|
| RMBG cutout | $0.0225 | $0.405 | Flat per image |
| Image upscale | $0.20 | $3.60 | Flat per image |
| Cutout plus upscale | $0.2225 | $4.005 | Two jobs |
| Nano Banana 2.1 at 4K | $0.20 | $3.60 | New image, not the same picture |
| GPT Image 2.5 high 4K | $0.2225 | $4.005 | New image, tiered price |
When to skip the upscale
An upscale makes an existing picture larger. A new 4K generation from Nano Banana 2.1 costs the same $0.20, but it makes a new picture. If you want to keep the exact photo you already approved, the upscale is the only route. If you just need a 4K hero and the composition is flexible, generating at the 4K tier costs the same and avoids a second job.
Check the pixel size your destination asks for. If your source already meets it, skip the upscale and keep the $3.60.
Cutout first or upscale first
Sume bills both as flat per-image jobs, so the order does not change the price. The order can change the result. Each order hands the second tool a different input, so the results can differ. Test both orders on two images before you run all 18: that is 2 x $0.2225 = $0.445 for each order, or $0.89 for both.
Hosted MCP lists both tools as paid and requires an idempotency_key on each call. Reuse the same key only for an exact retry, because the same key with a different payload is a conflict.
A decision table for 18 images
Most batches are not all-or-nothing. Imagine that 18 hero images split into 6 that need a transparent background only, 6 that are too small and need only an upscale, and 6 that need both. The cost is 12 cutouts (6 + 6) x $0.0225 = $0.27 plus 12 upscales (6 + 6) x $0.20 = $2.40, for $2.67 in total. That is $1.335 less than doing both steps on every image ($3.60 + $0.405 = $4.005).
The point of splitting is that the upscale dominates the bill. Every image you can exempt from the upscale saves $0.20, while every image you exempt from the cutout saves $0.0225. Inspect dimensions in the first pass, tag each file, and submit only the jobs that earn their price. Sume bills failed or cancelled generations nothing, but a completed job is charged in full even if you later discard the output.
Sources
Related posts
More in Media tools
- A 195-second render is too long for a Short: trim to 180 s for $0.02
Shorts take up to 3 minutes. Cut a 195 s Sume render to 180 s with video trim for a flat $0.02 instead of re-rendering the Timeline at $0.30.
- Lip sync on a 2.76:1 widescreen still: crop to 2.5:1 first
Sume's MiniMax H3 Max Lip Sync accepts stills with aspect ratio 0.4 to 2.5. A 2.76:1 frame is out of range; cropping 2760x1000 to 2500x1000 fits.
- A 300-second recording in 24 stills: video-frames fps 0.08
Video frames caps a call at 24 stills and a 300-second source. Send fps 0.08 and a 300-second clip gives 24 mid-bin stills, one every 12.5 seconds.
- 4:5 clip in Snapchat's 720x1280 frame: crop or pad 29.7%
A 1080x1350 clip in a 720x1280 Snap frame: cover loses 304 px of width, contain leaves 380 px of bars. Fit modes cover, contain, blur; $0.10 a minute.
Written by Sume