KefCull
Find your favorites. Keep control.
Turn a large photo collection into an organized selection of favorites. Local AI groups your photos by subject, swipe review helps you decide, and export copies your favorites without changing your originals.
A clearer collection.
Entirely yours.
Local AI. Original files preserved.
Everything you need to get going.
Automatic AI subject folders
Swipe right to keep, left to reject
Fullscreen review with subject overlays
Export favorites without changing originals
Search, sort, rate, and filter photos
Exact duplicate detection
A place for every subject.
Import a folder of JPEG, PNG, or WebP images. KefCull creates previews and suggests catalog folders such as People, Birds, Cars, Buildings, Flowers, Mountains, Food, and Landscapes.
- Open or rename a subject folder
- Browse its contents or review only those photos
- Your original files stay where they are
Keep. Reject. Come back later.
Review one photo at a time with mouse or touch gestures, on-screen buttons, or keyboard shortcuts. Fullscreen review gives your photo more room while preserving your place in the queue.
- Swipe right to keep; swipe left to reject
- Skip to decide later, or undo your last decision
- Adjust thumbnails, search filenames, sort, and add star ratings
- Filter kept, rejected, and unreviewed photos
- Inspect available dimensions, file size, camera, lens, and exposure information
A helping hand. You make the call.
DINOv2 Small extracts image features, TinyCLIP suggests subject categories, and YOLOX-Nano detects supported objects and their locations. The Subjects toggle shows labeled detection boxes during swipe review.
- Browse while analysis progress stays visible
- Pause and resume analysis; saved results are reused
- All model inference runs locally, without uploading photos
- Suggestions can be imperfect: no face recognition, photo-quality assessment, or automatic keep/reject decisions
- Bundled licenses and attribution are available in the app’s Model licenses screen
Install once. Review offline.
The Windows installer bundles the app and all three models. You do not need Python, Node.js, Rust, or a separate model download.
- Current release: v0.5.2 for Windows x64
- Supported images: JPEG, PNG, and WebP
- RAW, HEIC, and RAW+JPEG pairing are not currently supported
- Mobile apps and macOS distribution are not currently supported
- Dark interface with responsive controls and reduced-motion support
Your originals stay under your control.
Suggested folders organize your catalog; your original files stay in their existing locations. Keep and reject decisions are saved locally, and rejecting a photo does not delete it.
Export copies the original files, preserves their bytes, verifies the copies, and handles filename conflicts without overwriting existing files.
A few common questions.
How do I install KefCull?
Open the v0.5.2 GitHub release page, download KefCull-Setup.exe from Assets, and run it on Windows x64. The installer includes the app and its models; no separate runtime or model download is needed.
Which platforms and image formats are supported?
The current release supports Windows x64 and JPEG, PNG, and WebP images. Camera RAW, HEIC, RAW+JPEG pairing, mobile apps, and macOS distribution are not currently supported.
Does KefCull upload my photos or require online AI?
No. Photos, previews, analysis results, ratings, and review decisions remain on your computer. Model inference runs locally, and photos are not uploaded for analysis.
Will rejecting a photo delete the original?
No. Rejecting a photo records a local review decision. Your original files stay in their existing locations. Suggested subject folders organize the catalog rather than moving your originals.
A clearer collection starts here.
Get KefCull v0.5.2 for Windows x64. The installer includes the app and its models.