Bria Eraser API pricing: $0.0575 per image

Bria Eraser costs $0.0575 per image. Plan image volumes and budgets around your editing inputs.

Oct 28, 2026 6

Per image: $0.0575

TL;DR

Bria Eraser: Upload an image and a black-and-white mask to erase selected areas. No prompt is needed.
Each image costs $0.0575, with one PNG per request. View budget calculations
Budget $5.75 for 100 images. Check the edit on one set of inputs before scheduling a batch.

What does one image cost?

Bria Eraser costs $0.0575 per output image on HiAPI. Each request returns one result, so multiply the number of planned outputs by the unit price. Price checked on October 28, 2026.

Model	Per image	100 images	1,000 images
`bria-eraser`	$0.0575	$5.75	$57.50

The source image and mask define the edit. mask_type identifies the mask source, and preserve_alpha controls transparency handling. These settings do not form separate price tiers.

How much should you budget each month?

Planned outputs	Calculation	Image charges
100 images	100 × $0.0575	$5.75
1,000 images	1,000 × $0.0575	$57.50
10,000 images	10,000 × $0.0575	$575.00

Count every generated candidate when planning a budget, rather than counting only the images you eventually use. For example, three candidate images for each of 100 products means 300 outputs and $17.25 in image charges. The table covers the generation charges for those images.

What should you check before submitting?

Define the area to clean up before drawing a black-and-white mask. White covers the object to erase and black marks what to keep. Match the mask aspect ratio to the source image. For a transparent PNG, enable preserve_alpha and inspect the downloaded result’s alpha channel.

Eraser does not accept a prompt field. A sentence such as “remove the passer-by” cannot replace the mask. Use manual for a brush-drawn mask and automatic for a mask produced by a segmentation tool.

A source image and its result

Street photo: remove a passer-by

Open the model page to upload inputs, then upload the matching mask; no prompt is required.

A passer-by in a long coat occupies the center of the source image, and the white mask covers the person. In the result, paving fills that area while the buildings remain visible on both sides.

Cover the entire person in white and check the shoes and outer edges. Keep the tables and chairs you want to retain in the black area.

bria-eraser returns one image per request at $0.0575 per image.

Cleanup or a new product scene?

Task	HiAPI model	Per image
Erase a selected object with a mask	bria-eraser	$0.0575
Create a product scene from text or a reference	bria-product-shot	$0.0575

The models solve different tasks. Eraser needs a source image and mask. Product Shot needs a product photo plus a text scene or reference image. Start with your inputs and desired edit, then estimate the number of outputs.

Turn an output count into a working budget

List the source images and state the edit for each: a passer-by, tabletop clutter, a cable or a product sticker. If the same source needs different keep/remove decisions, those are separate editing plans. Count the candidates from every plan, not just the number of source files.

Budget the first pass separately from additional candidates. The first pass checks mask alignment, coverage and whether the filled area suits the intended use. Further candidates are useful once the inputs are correct and you still need to choose a fill. This shows whether the budget went into correcting inputs or selecting results.

The earlier example of 100 products with three candidates each is a planning scenario. It does not mean every product needs three attempts or that one will necessarily be usable. Review the first pass before allocating the remaining candidates to images that need work. Without your own acceptance history, do not assume an acceptance rate: track completed outputs and accepted images separately.

Which settings keep the same price, and what remains outside the estimate?

The current rate has no separate tier for manual versus automatic masks. preserve_alpha is not a paid upgrade either. Choose automatic because a segmentation tool produced the mask, and choose preserve_alpha=true because the input transparency needs to survive. Neither setting creates multiple outputs or replaces mask preparation and review.

The $0.0575 rate covers image generation. Plan your own labor for selecting sources, drawing masks, storing downloads and retouching separately. The multiplication here assigns no price to those activities. It also makes no refund promise for failures, cancellations or network interruptions; reconcile those cases against the actual task status and account billing.

For budget approval, record both the planned output count and the planned delivery count. The former drives the image-charge estimate; the latter drives the delivery schedule. A successful PNG result still needs an editing review before use in a product listing or campaign.

Which source images should go first?

Start with images whose target boundaries you can identify and mask, and confirm that the intended change is local removal. The street example removes a passer-by; the tabletop example removes paper and a tea bag. Moving a whole product into a new room calls for a different task. Enlarging the white mask does not describe a new scene.

Before a batch, check inputs representative of your actual work. A narrow cable and a broad sticker need different mask shapes; transparent assets also need an alpha-channel check. Passing the street example does not establish that your product stickers, edges or transparent regions will meet the delivery requirements.

When should you request another candidate or fix the input?

If part of an unwanted object remains, check mask coverage first. If the white area includes something you intended to keep, correct the selection. Both are input problems. If the mask is appropriate but the fill does not suit the delivery, another candidate may be useful. Record why you are trying again instead of repeatedly submitting a flawed mask.

Review the edited region and its neighbors together: paving continuity, the mug edge, wall tone, and the outline and alpha channel of a transparent product. Inspect the chosen result at its intended display size. More candidates cannot replace inspection of details that a thumbnail hides.