Backgrounds,Removed.

Verbier is a background removal API. Send it a photo and you get the subject back on a transparent background.

Get an API keySee pricing
  • 50 free images every month
  • No credit card required
  • Only delivered images count
  • Cancel any time

See the cut.

Drag across the photo to compare, and try the cut-out on a colour. These are the API's results exactly as they came back, on default settings and with no retouching.

  • Default settings, no retouching
  • Hair and fur keep their own colour
  • On a colour or transparent

JPEG, PNG or WebP up to 3 MB. Photos go through OpenAI's moderation check first. Verbier doesn't keep them; our privacy policy explains what our providers do.

Curly hair, background removed
Curly hair, original
⇆
BeforeAfter
The photo sent

POST /v1/cutout

verbier

The cutout returned

One request.

Post an image URL or base64 to /v1/cutout and a transparent WebP or PNG comes back, cropped to the subject, in about two seconds. The batch endpoint takes up to 32 images at a time, and every response tells you how much of your allowance is left.

  • Image URL or base64 in
  • Transparent WebP or PNG out, in about 2 s
  • Up to 32 images per batch call

Made for the hard ones.

Cut-outs usually go wrong on a white trainer against a white floor, a box cut off by the edge of the photo, or loose hair. Before cutting, Verbier pads each photo with more of its own backdrop, so a product that runs off the frame keeps its outline.

  • White on white
  • Products running off the frame
  • Hair, fur and thin stems
White sneakers on a white floor, background removed
Windblown red hair, background removed
Green frog on a green leaf, background removed
Dandelion seed head, background removed
A portrait lit by neon, shown on black. Cleanup removed 23 stray pieces, including the streak beside her arm.: "cleanup": false"cleanup": false
A portrait lit by neon, shown on black. Cleanup removed 23 stray pieces, including the streak beside her arm.: "cleanup": true"cleanup": true
A portrait lit by neon, shown on black. Cleanup removed 23 stray pieces, including the streak beside her arm.

No stray specks.

Turn on cleanup to remove the small bits of background a cut can leave behind, like a fleck of wall or a stray leaf. Second objects and loose strands of hair are left alone.

  • Removes stray specks of background
  • Keeps second objects and loose strands
  • "cleanup": true

Shadows that sit it down.

Ask for a shadow and Verbier draws it from the cut-out's own shape. A drop shadow lifts it off the page, a contact shadow darkens the ground where it stands, and a cast shadow falls back across the floor as it would under a studio light. The image grows to make room for it.

  • Drop, contact or cast shadow
  • Drawn from the cut-out's own shape
  • "shadow": "cast"
A vinyl collectible figure cut out with a cast shadow
The same portrait enlarged with upscale: 4
A portrait cut from a photo 500 px wide, enlarged four times with a plain resize
⇆
plain resize"upscale": 4
A photo 500 px wide, shown at four times its size. Drag to compare it with a plain resize.

Small photos, sharper.

Supplier photos are often tiny. Upscale makes the cut-out four times the size and keeps edges and texture sharp, where an ordinary resize would blur them.

  • 4× the resolution
  • Sharper than a plain resize
  • "upscale": 4

Trimmed to the subject.

Cut-outs come back cropped to the subject with a 2% margin, so there's no empty canvas to deal with. Set the margin to 0 for a tight crop, or turn trim off to keep the full frame, lined up with the original photo.

  • On by default, no extra cost
  • "margin": 0 for a tight crop
  • "trim": false keeps the full frame
A bicycle cut-out, cropped to the subject, with a 2% margin
803 × 485 px: cropped to the subject, with a 2% margin
White sneakers framed by the API: a white JPEG, the pair filling 90%, as Amazon asks
2000 × 2000 px: a white JPEG, the pair filling 90%, as Amazon asks

Framed for the marketplace.

One option frames the product for the shop it's going to. Amazon, eBay, Etsy, Shopify and Google Shopping each get a white JPEG at the size and fill that marketplace advises, and a small photo is never blown up: the canvas shrinks to fit it instead. Or choose your own canvas, backdrop and shadow.

  • Amazon, eBay, Etsy, Shopify, Google Shopping
  • Never blows up a small photo
  • "preset": "amazon"

Colour-matched to the scene.

Put the cut-out on a colour or on a photo of your own, and colour matching adjusts it to the light in the scene so it looks photographed there. It suits people and lifestyle shots. Leave it off when a product's colour has to stay exact.

  • Any colour or background photo
  • Lit to match the scene
  • "color_match": 0.4
The same dog with its colours matched to the room's warm light
A dog from a studio photo placed on a sofa in a warmly lit room, colours as shot
⇆
pasted in"color_match": 0.4
The living room the dog was matched to

Matched to this

A studio photo placed in a lamp-lit room. Drag to see it pick up the room's warm light.

Pay for what we deliver.

Every account gets 50 images a month free. Only images we send back count, so a failed request costs nothing. Upgrade or cancel from your dashboard.

  • 50 free images every month
  • Only delivered images count
  • Cancel any time

Background removal API questions.

What is a background removal API?

A web service that takes a photo and returns just the subject, with the background made transparent. With Verbier you post an image URL or base64 to /v1/cutout and get back a transparent WebP or PNG, cropped to the subject.

Is there a free background removal API?

Yes. Every Verbier account includes 50 images a month at no cost, and you don't need a card to sign up. Only images we deliver count towards it.

How much does background removal cost per image?

Starter is $9 a month for 1,000 images, about $0.009 each; Pro is $49 a month for 10,000 images, about $0.0049 each. Plans renew monthly and you can cancel from your dashboard at any time.

Can I remove backgrounds from many images at once?

Yes. The batch endpoint, /v1/batch, takes up to 32 images in one call and returns each result with the id you gave it. Each image counts once.

Which image formats does it accept and return?

Send JPEG, PNG, WebP, GIF, BMP or TIFF up to 20 MB, as a URL or as base64. You get back a transparent WebP by default, or a PNG if you ask for one. A JPEG comes back on white.

How well does it handle hair and fur?

Verbier cuts at high resolution, then cleans up the colour at the edges so strands keep their own colour instead of a fringe of the old background. The examples on this page are its results exactly as they came back.

Do you keep my images?

No. Verbier processes each image, returns the result and keeps no copy in its own systems, and never uses your images to train models. The provider that runs our processing servers deletes its encrypted copies within 7 days. Only counts and timings are logged, for usage and billing.

What happens if a request fails?

Nothing. An image only counts once it has been delivered, and every response carries your usage and limit in the X-Usage and X-Limit headers.

How fast is it?

About two seconds per image while it's running. After a quiet spell, the first request can take up to half a minute while a GPU starts up.

Can it add shadows, clean up or upscale?

Yes, as options on the same request. Every plan can use "upscale": 2 to double the resolution, and "margin" or "trim": false to control the crop. The paid plans add "shadow" ("drop", "contact" or "cast"), "cleanup": true to remove stray specks, "upscale": 4, "preset" to frame the product for Amazon, eBay, Etsy, Shopify or Google Shopping (or "canvas" for a size of your own), and "background" to put the result on a colour or photo, with "color_match" to adjust the subject to that background's light.

Sign up now

50 free images per month

Get an API key