Open Source · MIT · macOS · On-device
snapsift
**Keep the best shot. Lose the burst.**
Apple's Duplicates detector leaves manual bursts alone — hold the shutter, get fifteen near-identical frames, and it assumes you meant to. snapsift finds them, keeps the genuinely best one, and never touches a favorite. Built entirely on Apple's own frameworks, so nothing ever leaves your Mac.
How it works
1.
Scan
snapsift reads your library through PhotoKit and clusters the manual bursts — frames shot seconds apart at the same size — that the Duplicates album never flags. A second neural pass can also find the same photo saved across different days.
2.
Review
Each cluster is laid out side by side with the keeper highlighted. snapsift picks it using Apple's own quality scores and on-device face analysis — sharpest frame, eyes open. Click any frame to keep that one instead; favorites are locked.
3.
Delete
Deleting moves the extras into Recently Deleted through Photos itself — recoverable for 30 days and synced across your devices. No AppleScript, no irreversible writes, no surprises.
What's inside
Fully on-device
PhotoKit reads, Vision analyses, and deletion goes through Photos. Nothing is uploaded, nothing is tracked — your library never leaves your Mac.
Keeps the genuinely best frame
Reads Apple's own per-photo quality scores and runs Vision face + open-eyes analysis, so the survivor is the sharpest, best-composed frame where everyone looks their best — not just the biggest file.
Cross-time look-alikes
Apple's neural feature prints find the same photo saved on different days — re-downloaded, screenshotted, AirDropped back — not just consecutive bursts.
Safe by default
Favorites are never deleted, videos are skipped unless you opt in, and every deletion is recoverable for 30 days. The aggressive options exist, but you turn them on.
Your library stays yours
Free and open source, built on the frameworks already in your Mac. snapsift reclaims the gigabytes wasted on near-duplicates without sending a single photo to anyone — and ships as both a native app and a hackable Python engine. Build it from source with the Xcode Command Line Tools — no full Xcode, no cloud, no cost.