Remove background from images in bulk with an API

To remove the background from images in bulk, submit one background-removal job per image and let the jobs run side by side. Sume's RMBG 1.0 takes a single image_url per request and returns a PNG with an alpha channel, at one price per image whatever its size, so a catalog of 1,000 product photos is 1,000 jobs, paced by your workspace's concurrency limit.

RMBG 1.0 is specified in the OpenAPI document behind the Sume API reference; the queue and job rules come from the Generation admission and Jobs and results docs, all read on 2026-09-27. The single-image call is covered in Remove background API.

How do I remove backgrounds from many images at once?

The request takes one public HTTPS image_url, so every photo must already be at a public HTTPS URL before the batch starts. Then loop over the list:

• Send one POST /v1/rmbg-1.0/remove per image, with an Idempotency-Key derived from the image, such as its SKU. A retry with the same key and body returns the original job instead of a second paid one.
• Use mode: "async", the default, or mode: "webhook" with a webhook_url. Both return the job id at once.
• Keep a map from each image to its job id, and fetch GET /v1/jobs/{id}/result once result_ready is true. Before that it answers 409 job_not_completed.
• Completed results expose mirrored PNG artifacts with alpha. Download them to your own storage.
• The docs do not describe how edges such as hair or glass come out, so run a sample before the whole catalog.

curl -X POST https://api.sume.com/v1/rmbg-1.0/remove \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: rmbg-sku-4411-v1" \
  -d '{
    "image_url": "https://example.com/catalog/sku-4411.jpg",
    "mode": "webhook",
    "webhook_url": "https://example.com/webhooks/sume"
  }'

Should I poll each cutout or use a webhook?

For a batch, pick a mode that returns at once; every mode returns the job id in the first response. A webhook URL must be public HTTPS, since localhost, private-network, and non-HTTPS URLs are rejected, and a job.completed payload carries the artifacts.

A webhook is a delivery optimization, not your only recovery path, so keep status_url polling for missed deliveries. Verifying the signature is covered in signed webhooks. The three modes compare like this:

How many cutouts run at the same time?

In current code each request creates a background_removal job, which goes through the same queue-first admission as other paid generation jobs. Your workspace's concurrency_limit sets how many process at once, queued jobs wait for a free slot, and when the queue is also full the submit fails with 429 queue_full. On queue_full, stop adding work and wait for jobs to finish.

The pacing rules are the same as for any image batch, covered in bulk image generation; which calls take a concurrency slot lists the other job types.

How much does bulk background removal cost?

API pricing lists background removal at $0.0225 per image, plus a 5.5% agent fee by default, and the public catalog adds that the price does not vary by image size. 1,000 cutouts therefore come to $22.50 before the agent fee.

Sume reserves each job's estimate when it accepts the submit, captures it on success, and refunds it on failure or on a cancellation before capture. Cancellation works only before generation starts.

Can I upscale images in bulk the same way?

Yes. POST /v1/image-upscale-1.0/upscale also takes one public HTTPS image_url per job, so the same loop, keys, and pacing apply, and API pricing lists it at $0.20 per image. Neither endpoint takes a list of images: one URL, one job. The upscale factor and output formats are in AI image upscaler API.