SINGLE + BATCH
A single image or a full folder.
Drag in a file, browse, or paste an image. In batch mode, select multiple images or a folder, including its subfolders. Process the selection and export everything together.
Thoughtfully simple
Everything you need.
Room to create.
Remove the background. Refine the result. Export it your way.
A focused workflow, with your images always on your computer.
EDGE REFINEMENT
Find the edge that feels right.
Make cutout edges sharper or softer with the Edges slider. A live preview helps you see your adjustments before saving.
BACKGROUND OPTIONS
A fresh setting for your subject.
Export on transparency, put a color behind the subject using the color wheel or a hex value, or choose a picture as the background. Background options work for batches, too.
FLEXIBLE EXPORTS
Your image, at its original size.
Export PNG, JPG, or WebP at the original dimensions. Choose PNG for transparency. JPG uses a background because the format cannot store transparency.
ZOOM + PAN
Get a closer look.
Zoom and pan the original and processed previews together. Reset the view when you’re done inspecting fine details.
EVERYDAY DETAILS
Make yourself at home.
Choose Dark, Light, or System appearance. CutOut remembers its window size and position. Cancel processing at any time, including with Esc.
What’s behind the cut
Four models.
One local workflow.
Different images ask for different approaches. All four models are bundled in CutOut; switch between them without downloading anything extra.
Fast
U²-Net-P
The lightest option, using a 320 × 320 model input. Start here when turnaround matters more than fine edges.
Apache-2.0Balanced
IS-Net general use
A general-purpose model with a 1024 × 1024 input. A useful balance for everyday image work.
Apache-2.0Best
BiRefNet Lite
Uses a 1024 × 1024 input, with a focus on fine detail such as hair. Requires more processing time than the lighter options.
MITBest+
BiRefNet · fp16
The full BiRefNet model in fp16, with a 1024 × 1024 input. The repository describes this as its slowest, highest-quality option.
MITThese are model choices, not measured quality scores. Results and runtime depend on the image and computer. Internal model input sizes do not limit the exported image dimensions. The repository recommends 8 GB RAM or more, and the Best models can briefly use several GB.
Model licenses and creditsA little less friction
Your images.
Your computer.
Your creative space.
Right here. On your computer.
AI processing happens locally. Open an image and get to work, even without an internet connection.
Your images stay your images.
No uploading to a server. No account to create. Your photos stay on the device you choose.
Free, and open to everyone.
No subscriptions or paywalls. The code is open source under the MIT license.
One image. Or the whole folder.
Make a quick cutout, process a selection, or work through a folder and its subfolders in one batch.
Make room for the subject
Ready for
a clean start?
Get CutOut. Keep your images local.
Let your ideas go a little further.
MIT licensedNo accountWorks offline
CutOut for Windows
v0.1.1
x64 .exe698 MB · Models included
Download .exeWindows 10 / 11 · 64-bit · AVX2 CPU required. 8 GB RAM recommended. The installer is not code-signed.
Showing the verified release snapshot from Oct 19, 2026.