Background removal at $0.0225 per image, flat for any photo size
Sume's background removal costs $0.0225 per image and the catalog says the price does not vary by image size. 2,000 photos cost $45; 40,000 cost $900.
Sume's Background Removal (sume/rmbg-1.0) costs $0.0225 per image, and the catalog row states that the price does not vary by image size. A 500-pixel thumbnail and a large studio photo cost the same, so the bill is the image count times $0.0225: $45.00 for 2,000 photos and $900.00 for 40,000.
The price table
The catalog describes one background-removal generation as $0.0225 after Sume's margin (read 2026-10-09). There is no per-megapixel tier to budget for.
| Images | Arithmetic | Cost |
|---|---|---|
| 500 | 500 x $0.0225 | $11.25 |
| 2,000 | 2,000 x $0.0225 | $45.00 |
| 12,500 | 12,500 x $0.0225 | $281.25 |
| 40,000 | 40,000 x $0.0225 | $900.00 |
What a flat price changes
With a size-based price you would downscale before sending to save money. With a flat price there is no saving in doing so, so send the original and keep full resolution in the cutout. If you want a smaller file, resize the result afterward, which costs nothing on Sume.
The call is a job. POST /v1/rmbg-1.0/remove returns a job and you read GET /v1/jobs/:id/status and GET /v1/jobs/:id/result; the catalog also lists the model-run alias POST /v1/models/sume/rmbg-1.0/runs. The reservation rounds up to whole cents at submit (3 cents for this model) and the captured amount is the Sume price.
The catalog lists the row as a fixed estimate, minimum and maximum of the same value, so a wallet check is exact: N images need N x $0.0225, with the reservation rounded up per job.
Background removal versus asking an image model
Some image models take background: transparent at generation time, which GPT Image 2.5 documents on Sume. That is for new images. If you start with a photo, the removal model is the route that leaves the original pixels alone. Compare cost per cutout: removal is $0.0225, while the cheapest GPT Image 2.5 row (low, 1K) is $0.02475, so removal is cheaper and does not regenerate the subject.
Planning a catalog job
For a catalog of 12,500 photos the cost is 12,500 x $0.0225 = $281.25, and the catalog row shows 3 cents as the per-job reservation, so a burst of 100 jobs holds about $3.00 until they complete. If you submit in waves of 100, keep the wallet above the wave plus what you expect to be captured.
Retry only the jobs that failed. A failed job is not billed, so resubmitting the same image costs nothing extra, and an Idempotency-Key header stops a client retry from creating a duplicate job when the original actually succeeded. Use the same key only for the same operation and payload.
Sources
Related posts
More in Media tools
- Best of 8 music takes costs $1.00 on Sume, then a 1-cent cut
Music 1.0 and the Music Router charge a flat $0.125 per track and reject a duration field. Generate 8 takes for $1.00, then cut the best one with a $0.01 split.
- Talking photo greeting: 8 seconds on Fabric 720p plus voice-over cost
An 8-second talking-photo greeting on VEED Fabric 1.0 costs $1.50 at 720p or $0.80 at 480p, plus about $0.0114 of TTS for a 240-character line.
- Korean captions API error: caption_hangul_text_latin_style (400)
Sume refuses Korean text on the slam, punch and tiktok-green caption styles with a 400. Use black-outline, korean-ad or another Hangul style; $0.20 a job.
- caption_no_speech: burn authored cues onto a silent clip, no ASR
A silent clip makes speech-to-captions fail with caption_no_speech. Send cues (text, start, end) and Sume burns your text at those times without speech-to-text.
Written by Sume