Remove image backgrounds on your Mac. Nothing leaves it.

What it does

Open a photo. Get the subject back.

Open a photo or drag one onto the window, and Transparentify finds the subject for you, whether it is a person, a pet, a product, a car, or a logo. Press Process and the result appears next to the original on a checkerboard, so you can compare the two before you save anything. When you are happy with it, save it as a PNG, copy it to the clipboard, or drag it straight into another app. If you would rather have a solid colour behind the subject than transparency, you can choose one instead.

The cut-out is made by an image model that is built into the app and runs on your Mac. It is designed for difficult edges: hair and fur keep their fine strands with real partial transparency, and Transparentify removes the colour of the old background from the edges, so the subject does not carry a halo when you place it on something new. When the original background is flat, soft details such as glows, shadows, and feathered outlines are kept as well. Images that already have transparency keep it, and two sliders let you tighten or soften the outline.

When the first result is not quite right, you can correct it. Draw a box around the subject you want and Transparentify looks again inside that box only, which helps when part of the subject was missed or when unwanted objects came along. The touch-up brush lets you paint back anything that was cut away or erase anything that should not be there, and every stroke can be undone. When a photo contains several subjects, each one is shown separately and you choose which ones stay. You can zoom in and pan around to check fine detail at full resolution.

Have a folder of product shots? Open the whole folder and Transparentify processes every image in it, saving a PNG next to each original.

Features

Everything you need for a clean cutout.

Under the hood

An open model, running on your Mac.

Transparentify finds the subject with BiRefNet HR, an open-source image segmentation model by Peng Zheng and colleagues, released under the MIT licence. The app ships with the model converted to Core ML, so it runs on your Mac's GPU, typically in under a second per image on Apple silicon, and never needs a connection.

The model's mask is then cleaned up by the app's own processing. Faint, separate patches that are not part of the subject are dropped. Edge pixels whose colour matches the nearby background have their opacity lowered, which removes the fringe that often surrounds hair and fur, and the remaining semi-transparent pixels have the background's colour taken out of them. When the background is flat, a separate step recovers soft details such as glows and shadows with true partial transparency.

Apple's Vision framework is used alongside the model to tell separate subjects apart, which is what lets you choose which ones stay. Core Image and Accelerate handle the image processing and final compositing. Because the model is part of the app, your images never leave your Mac.

Known limits

Stated rather than hidden.

• Objects that touch a person, such as a chair they sit on, are sometimes left out. Draw a box around everything you want to keep.
• Flat panels attached to a subject, such as a sign on a post, may be removed with the background.
• On small or low-resolution images, results can change noticeably when the box is moved slightly.
• Very fine hair against a busy background can still have simplified edges.
• Output is PNG, 8 bits per channel.

Requirements

What you need.