# CrystalCut

> Fast images. Private processing.

AI background removal, object selection, brush refinement and batch conversion in one workspace.

- Developer: CommonSource
- Platforms: Windows, macOS
- License: Apache-2.0
- Official product page: https://commonsourcelab.com/en/desktop/crystalcut

## When one click is not enough

Inspect the automatic result and fix only what needs attention. Use the same flow for one image or a whole folder.

01. **Remove**: AI separates the subject first.
02. **Review**: Compare the source and result.
03. **Refine**: Brush only where correction is needed.

## Key features

- **Refine only what needs attention**: Use Keep and Remove brushes over the automatic result to control exactly what stays.
- **Compare source and result**: Switch among original, preview, mask and compare views before saving.
- **One image or a whole folder**: Review a queue, choose format, size, naming and destination, then export in a batch.
- **Local processing, visible settings**: Images stay on your computer. Metadata, GPS and generation details are included only when you choose.

## Privacy and processing

Background removal, previews, refinement and encoding run on your device. Only the model required for AI features is downloaded on first use.

### Open down to the code

The React interface and Rust image engine are available under Apache-2.0. Images never leave your computer.

## Stores

- [Microsoft Store](https://apps.microsoft.com/detail/9n3jr8npmq92)
- Mac App Store: Coming soon

## FAQ

### Which files can CrystalCut process?

Import JPEG, PNG and WebP, then export PNG or lossy or lossless WebP. Convert-only mode can keep the background intact.

### Are images uploaded to a server?

No. Removal, previews, masks, resizing, metadata handling and encoding run locally. Only AI model files are downloaded on first use.

### Where will the app be available?

The Windows version is available from Microsoft Store. The Mac App Store version is in preparation and will be linked here after release.

### Can it be open source and sold in stores?

Yes. The code is published under Apache-2.0, while official stores will offer paid packages for convenient installation and updates.