RMBG, image upscale and video upscale request fields side by side

Sume's three cleanup tools take three different request bodies. Remove-background needs only image_url. Image upscale adds upscale_factor (1 to 4) and output_format. Video upscale uses video_url with upscale_factor from 1.1 to 4, duration_seconds from 1 to 30 and an enhancement_tier. This page puts them in one table with the prices.

The fields

Each route is a POST that returns a job. They all accept the usual communication fields (mode, webhook_url, wait_timeout_seconds) and an optional metadata object, which carries your own record such as a SKU. None accepts model, endpoint or provider_endpoint. Unknown fields are rejected.

Notes on each

For video upscale, scale_ratio is a second name for the multiplier. If you send both, scale_ratio wins, as the post on which field wins shows, so send just one. Omit duration_seconds and Sume reserves 5 seconds, so set it to the real length.

For image upscale, the catalog lists a single $0.20 price, and its description says the hold covers about 16 megapixels of output. For remove-background, the result is a PNG with an alpha channel.

Using them from an agent

The hosted MCP server exposes the same tools as rmbg_create, image_upscale_create and video_upscale_create. They are paid tools, so each call carries an idempotency_key. Read the job with jobs_wait and jobs_result afterward.

Use metadata on every request. When 500 jobs return in an arbitrary order, it is the cheapest way to match a result to a SKU without a lookup. Keep your own ledger as well, since metadata is a convenience and not an audit log.

Example bodies

The smallest valid body for each tool is one URL. Add the optional fields only when you need them, since each one narrows what Sume can do for you. A body like the ones below is the whole request. The key point for budgeting is that none of these fields changes the price of an image tool, while the duration changes the video price.

The video example sets a factor of 2, a 12-second duration and the standard tier. Cost: 12 x $0.009 = $0.108.

{"image_url": "https://example.com/a.jpg"}

{"image_url": "https://example.com/a.png", "upscale_factor": 2, "output_format": "png"}

{"video_url": "https://example.com/a.mp4", "upscale_factor": 2,
 "duration_seconds": 12, "enhancement_tier": "standard"}

Choosing a tier

enhancement_tier takes fast, standard or pro. The catalog lists one per-second price for the tool, $0.009, and does not publish a different price by tier. Test the tiers on one clip each and compare the outputs, because the catalog price will not tell you the difference.