25 tools for product photos.

An AI model finds the product, and the old background’s colour is cleaned out of its soft edges, so a bottle shot against a green wall keeps no green rim. Keep the transparency, or put the product on white, a colour, an image of yours or its own photo blurred. Crop to it with a margin, or place it on a canvas of any size, standing or centred, with a contact or drop shadow and a reflection.

Examples and more on Remove background

You choose

• Background: Transparent, White, Colour, Its own photo, blurred, An image of yours
• Blur
• Background image
• Size: As the photo, Cropped, Canvas
• Margin around it
• Canvas width
• Canvas height
• The product fills: 95%, 85%, 75%, 60%, 50%
• Place it: Centred, Standing, Hanging
• Shadow: None, Contact, Drop
• A reflection under it

It takes

One photo at a time, or a whole folder at once in the browser.

You get

A PNG with transparency, or a JPEG on white or your colour.

Cost

1 credit a photo

In the API

POST /v1/remove-background

Every line of text in the photo is found, and the lines off the product are painted out: watermarks, captions, prices and shop names. The product’s own label stays. You can also remove text on the product, or only the lines that match words you give.

Examples and more on Remove text and watermarks

You choose

• Remove text: Off the product, Everywhere
• Only text matching

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo without the text.

Cost

1 credit a photo

In the API

POST /v1/remove-text

The fill comes from an AI model made to rebuild what a removed object hid, from the texture and light around it. Separate areas are filled one at a time, each at full detail.

Examples and more on Erase objects

It takes

One photo, and the area you paint over it.

You get

The photo with the areas filled in.

Cost

1 credit a photo

In the API

POST /v1/erase

You say what the subject is: click the thing you want, click again to add to it, and mark what must stay out. An AI model answers a click three ways, as the part, the object and the whole it belongs to, and the one it is surest of that agrees with every click is cut out, with its own colour along the soft edge. Or take its mask. Edges are softer than the background remover’s on large photos, and hair is not followed.

Examples and more on Select an object

You choose

• What a click means: The likeliest, The part, The whole thing
• Result: Cut out, transparent, Cut out, on white, Its mask
• Crop to it

It takes

One photo, and a click on the thing you want in it.

You get

The object cut out, or its mask, and how sure the model is of the fit.

Cost

Free

In the API

POST /v1/select

Without settings, the tones are stretched to the full range and a colour cast is taken out, with one gain for all three colours so nothing changes hue. Or set brightness, contrast, saturation, temperature and sharpening yourself. It is not an AI enhancer: it does not remove noise or blur, or recover blown highlights.

Examples and more on Fix light and colour

You choose

• Automatic levels and white balance: On, Off
• Brightness
• Contrast
• Saturation
• Temperature
• Sharpen

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo with its light and colour fixed.

Cost

1 credit a photo

In the API

POST /v1/enhance

A face detector finds every face, and each one is hidden inside the oval around it: blurred, pixelated, or covered with a colour. Blurring and pixelating stop a person from recognising the face; covering leaves nothing of it. A face must be about 1/60 of the photo’s longest side to be found, and a head seen from behind is not a face: check the count of faces hidden.

Examples and more on Hide faces

You choose

• Hide them: Blur, Pixelate, Cover
• Strength
• Room around each face: 0%, 15%, 30%, 50%
• Faces to find: Only clear ones, As usual, More, some not faces

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo with the faces hidden, and how many were.

Cost

1 credit a photo

In the API

POST /v1/blur-faces

An AI model removes sensor noise from a dark or high-ISO shot, and the blocks and ringing of heavy JPEG compression, deciding for itself how much there is. Strength blends the cleaned photo back toward yours. It does not sharpen a blurred photo or add detail; on a photo that is already clean it smooths fine texture. Up to 12 megapixels.

Examples and more on Remove noise

You choose

• Strength

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo without its noise.

Cost

1 credit a photo

In the API

POST /v1/denoise

An AI model predicts colour from a small copy of the photo, and the photo keeps its own light and detail at full size. The colours are guesses from what such things usually look like: skin, sky, grass and wood come out believable, while the colour of clothes, cars and walls cannot be known and is invented. A photo that already has colour is treated as black and white.

Examples and more on Colourise

You choose

• Colour

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo in colour.

Cost

Free

In the API

POST /v1/colorize

A face detector finds each face, and an AI model redraws it sharp and puts it back with a soft edge; nothing else in the photo changes. For old, blurred, small or noisy photos. The face is drawn from what faces look like: it resembles the person, but details such as the colour of the eyes can differ from the truth, so do not use it for an identity document or as evidence. A face that is already large and sharp is left alone.

Examples and more on Restore faces

You choose

• Strength
• Faces to find: Only clear ones, As usual, More, some not faces

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo with its faces redrawn, and how many were.

Cost

Free

In the API

POST /v1/restore-faces

An AI model enlarges a small photo two, three or four times, and draws edges and text sharp instead of blurring them. Use it when a photo is under a marketplace’s minimum size.

Examples and more on Upscale

You choose

• Enlarge: 2×, 3×, 4×

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo, two to four times larger.

Cost

1 credit a photo

In the API

POST /v1/upscale

The background-removal model finds the product, or a face detector the faces, and the photo is cropped to the shape you choose: the largest crop of that shape, moved so the product is in the middle, or a close crop with room around it. The background stays.

Examples and more on Crop to a shape

You choose

• Shape: Square, 1:1, Portrait, 4:5, Portrait, 3:4, Portrait, 2:3, Landscape, 4:3, Landscape, 3:2, Wide, 16:9, Story, 9:16
• Around: The product, The faces
• Crop close around it
• Room around it: 5%, 10%, 20%, 30%, 60%, 100%

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo cropped to the shape, and whether a product was found.

Cost

1 credit a photo

In the API

POST /v1/crop

The photo is put on a larger canvas, to a shape or by pixels on each side, and an inpainting model continues it into the new area. That suits a studio backdrop, a table, a wall or sky. It does not invent content: a product cut off at the edge stays cut off. Or fill the new area with white or a colour.

Examples and more on Expand the canvas

You choose

• Add room: To a shape, By pixels
• Shape: Square, 1:1, Portrait, 4:5, Portrait, 3:4, Portrait, 2:3, Landscape, 4:3, Landscape, 3:2, Wide, 16:9, Story, 9:16
• Top
• Bottom
• Left
• Right
• The new area: Continued, White, Colour

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo on its larger canvas.

Cost

1 credit a photo

In the API

POST /v1/expand

Give a width, a height or both, and fit the photo inside, fill and crop, or stretch. Trim an even border first, convert to JPEG, WebP, PNG or AVIF, and give a file size: the quality is lowered until the photo fits, for shops and marketplaces with upload limits. Set a print resolution for a JPEG or PNG.

Examples and more on Resize and compress

You choose

• Trim an even border first
• Border tolerance
• Width
• Height
• With both sides given: Fit inside, Fill and crop, Stretch
• Enlarge smaller photos to the size
• Format: Same as the photo, JPEG, WebP, PNG, AVIF
• File size at most: No limit, 100 KB, 300 KB, 500 KB, 1 MB, 2 MB
• Quality
• Print resolution

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo at its new size, in the format you chose.

Cost

1 credit a photo

In the API

POST /v1/resize

Your logo, with its transparency, or a line of text in a bold sans-serif, placed in one of nine positions or repeated across the photo. Its size is a share of the photo’s width, so photos of any size are marked alike.

Examples and more on Add a watermark

You choose

• Mark: Text, Logo
• Text
• Logo
• Repeat it across the photo
• Position: Top left, Top, Top right, Left, Centre, Right, Bottom left, Bottom, Bottom right
• Size
• Opacity
• Margin

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo with your mark on it.

Cost

1 credit a photo

In the API

POST /v1/watermark

The one face is found, and the person’s outline gives the top of the head. The photo is turned so the eyes are level, then scaled and cropped so the head, chin to crown, sits in the middle of the document’s range, on a photo of its size at 300 or 600 DPI. Built in: the passports of the United States, the United Kingdom, Canada and Australia, or give any document’s sizes. The background is kept unless you replace it. What it sees wrong is listed (a tilted head, a face turned away, a head cut off, too little room, low resolution, a busy background); expression, glasses, shadows and likeness are for you to check.

Examples and more on Passport photo

You choose

• Document: United States passport (2 × 2 in), United Kingdom passport (35 × 45 mm), Canada passport (50 × 70 mm), Australia passport (35 × 45 mm), Another: give its sizes
• Photo width
• Photo height
• Head at least, chin to crown
• Head at most
• Background: Keep it, As asked, White
• Background: Keep it, White
• Print resolution: 300 DPI, 600 DPI

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The photo at the document’s size, the head’s height in millimetres, and what was found wrong.

Cost

1 credit a photo

In the API

POST /v1/id-photo

Each preset runs several tools in one step. Marketplace: pure white, square, the product filling 85%, a JPEG under 2 MB. Studio: a soft grey gradient with a contact shadow. Social: a 4:5 post over a blurred copy of the photo. Cut-out: a transparent PNG cropped to the product. Web: 1600 pixels on the long side, a WebP under 300 KB.

Examples and more on Presets

You choose

• Preset: Marketplace, Studio, Social, Cut-out, Web

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The finished photo.

Cost

1 credit a photo

In the API

POST /v1/run

Each area of one colour becomes a filled shape, traced with smooth curves, straight lines or pixel steps. Made for logos, icons and flat drawings; a photograph traces into thousands of shapes and a large file. Transparent areas stay empty.

Examples and more on Trace into SVG

You choose

• Colours: In colour, Black only
• Lines: Smooth, Straight, Pixels
• Colour detail
• Drop specks smaller than

It takes

One photo at a time, or a whole folder at once in the browser.

You get

An SVG file, and how many shapes it has.

Cost

1 credit a photo

In the API

POST /v1/vectorize

The marketplace’s rules decide what is done. Where white is required, the product is cut out onto pure white and framed at the marketplace’s size and fill. Where added text is not allowed, it is painted out. Every photo is sized and compressed to the limits, and a small one is upscaled only when it is under the marketplace’s minimum. Then the result is checked again.

Examples and more on Fix for a marketplace

You choose

• Marketplace: Amazon (main image), eBay, Etsy, Shopify, Google Merchant Center, Allegro, Zalando (packshot)

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The fixed photo, and whether it now passes.

Cost

1 credit a photo

In the API

POST /v1/fix

Sharpness, noise, exposure, colour tint and framing are measured at the size a shopper zooms to. With a marketplace chosen, its rules are checked too: what it rejects is an error, what it only recommends is a warning, and what a machine can’t judge, like props or packaging, is listed for you to look at.

Examples and more on Check photos

You choose

• Check against: Photo quality only, Amazon (main image), eBay, Etsy, Shopify, Google Merchant Center, Allegro, Zalando (packshot)
• Find the product: framing and background
• Read text in the photo

It takes

One photo at a time, or a whole folder at once in the browser.

You get

A report for each photo: what was measured, what was found, and whether it passes.

Cost

1 credit a photo

In the API

POST /v1/check

Photos of one product are read together: barcodes and their GS1 data, net quantity, dates, lot, ingredients, EU allergens, nutrition per 100 g, alcohol, age and vintage, fabric, voltage and power. Each field keeps the text it came from. Labels in English, German, French, Italian, Spanish, Dutch and Croatian.

Examples and more on Read a label

It takes

Up to 8 photos of one product, read together.

You get

The label’s facts as fields, each with the text it came from.

Cost

1 credit a photo

In the API

POST /v1/label

EAN-13, UPC-A and ITF-14 codes are found and their check digits verified, so a misread code is never returned. Ask for QR, Data Matrix, Code 128 and other formats too.

Examples and more on Read barcodes

You choose

• Also QR, Data Matrix, Code 128 and other formats

It takes

One photo at a time, or a whole folder at once in the browser.

You get

Each code’s format and value, and the GS1 data it carries.

Cost

1 credit a photo

In the API

POST /v1/barcodes

Every line of text in the photo, top to bottom, read by an OCR model that knows many languages, with where each line is and how sure the reading is.

Examples and more on Read text

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The lines of text, each with its position and confidence.

Cost

1 credit a photo

In the API

POST /v1/ocr

A face detector finds every face, with a box from forehead to chin, the eyes, nose and mouth corners, how sure it is and how far the head is tilted. It finds faces; it does not recognise whose they are, and nothing is compared or kept.

Examples and more on Find faces

You choose

• Faces to find: Only clear ones, As usual, More, some not faces

It takes

One photo at a time, or a whole folder at once in the browser.

You get

Each face’s box, landmarks, score and tilt.

Cost

1 credit a photo

In the API

POST /v1/faces

Up to 16 colours, the most common first, each with its share of the photo. Transparent areas do not count, so a cut-out gives the product’s own colours: fill in a listing’s colour, or match a background to the product.

Examples and more on Find the colours

You choose

• Colours at most

It takes

One photo at a time, or a whole folder at once in the browser.

You get

Each colour’s hex code, its RGB values and its share of the photo.

Cost

1 credit a photo

In the API

POST /v1/palette

An AI model describes what it sees in English: one sentence for a caption or alt text, a few, or a paragraph. The same photo always gets the same words. It can be wrong about a colour, a count or what a thing is, more often in the longer forms, and it knows no names, brands or places unless they are written in the photo: read it before you publish it.

Examples and more on Describe the photo

You choose

• How much to say: One sentence, A few sentences, A paragraph

It takes

One photo at a time, or a whole folder at once in the browser.

You get

The description, as text.

Cost

Free

In the API

POST /v1/caption

Upload your products as a spreadsheet: an id and a title for each, and the brand, barcode and SKU where you have them. The photos are read for their barcodes and the text on the label, and your products are ranked by how much of that agrees with them: a barcode or an SKU settles it, then the brand and the words of the name. Up to 8 photos of one product are read together, so the front, the back and the barcode count as one. A food or cosmetics catalogue also compares the net quantity, so the right size of a product sold in several is found, and the multipack or the SPF; a pharmacy catalogue the strength and the number of tablets in the pack; a household or pet catalogue the washes, rolls or wipes; an electronics catalogue the capacity, the model number and the batteries in the pack; a wine or spirits catalogue the age, vintage, strength, size and cask number. It reads; it does not know a product by its look, so one with no text or barcode on it cannot be matched.

Examples and more on Match to your catalogue

You choose

• Catalogue

It takes

Up to 8 photos of one product, read together.

You get

Your products that fit best, each with its id and title as in your catalogue, a score to rank them by, and what agreed, such as the barcode, the SKU or the brand.

Cost

1 credit a photo

In the API

POST /v1/identify