Requirements and compatibility
Check practical hardware guidance and distinguish automated test environments from desktop installations still awaiting validation.
Guide based on source version 2.0.1-rc.7. Screenshots show the English interface of a prerelease edition; their values illustrate examples, not necessarily defaults. Packages on Downloads may differ: review their release notes and saving options before processing.
On this page
Hardware for the full app
This guidance includes OCR and Remove Background, as well as the regular tools.
| Component | Provisional minimum | Recommended |
|---|---|---|
| Processor | Relatively recent Intel/AMD x86-64, 4 cores | Modern 6–8-core CPU |
| System RAM | 8 GB, with few other apps open | 16 GB; 32 GB for very large images and multitasking |
| Storage | 4 GB free, excluding originals and results | SSD with 10 GB free or more, excluding images |
| Graphics | Integrated graphics; no mandatory dedicated GPU | Integrated graphics |
| Internet | Initial model downloads and update checks | Offline work once the required models are downloaded |
These are practical estimates, not validated hardware minimums. Low-end PC testing is still pending. A 4 GB RAM computer is not recommended for full OCR and Remove Background use. Resolution, file counts, and loaded models greatly affect consumption.
The tested package uses CPU: it does not require an NVIDIA or other dedicated GPU. AI libraries do not guarantee GPU acceleration in every installation.
Systems and architecture
| System | Reference status |
|---|---|
| 64-bit Windows 10/11 | Windows 11 preferred; final clean-machine package acceptance is pending on both systems. |
| Ubuntu 22.04 x86-64 | CI baseline. Other desktops and distributions require testing the final AppImage. |
| macOS | Experimental; no validated native package is declared. |
| ARM | No validated native version of the current packages. |
ONNX Runtime lists Windows 10 as a Windows minimum in its inference documentation. That is a dependency requirement, not certification of the complete app on every machine.
CI uses Windows Server and Ubuntu 22.04 with a virtual display. It does not replace manual checks of installation, scaling, clipboard, fonts, and desktop behavior.
Disk space and models
Besides the program, allow space for originals, results, OCR and background models, caches, and possible temporary AppImage extraction. A source environment usually takes more space than a distributed package.
Historical tests on October 15, 2026 observed approximately 3 GB of process RAM on Windows and 2.6 GB on Linux with several models loaded. Measured app, model, and cache storage was 1.10 GB and 1.53 GB. First ISNet initialization took about 53 and 39 seconds respectively. These are observations from those tests, not maximums or guaranteed timings for this version or your PC. They are not per-photo processing or download times.
Formats and running from source
File selectors and codecs vary by module. A format appearing in a dialog does not guarantee every operation can read or write it. Check Formats and limits.
Running from source also requires Python 3.13, Tk, the appropriate dependency file, and platform libraries. Use the Windows or Linux instructions instead of treating the prebuilt package requirements as a complete development environment.