What you already told Photos
Favourited, edited, cropped, filed in an album, shared. These outrank every algorithm in the app. If you touched a frame, it wins its group — and Keeper says that’s why.
Coming soon · in design
Keeper
“iPhone Storage Full” is never the screenshots you meant to keep — it’s the sixteen-frame burst you never culled, the screen recording from two years ago, and 4,000 receipts. Keeper finds all of it, picks the best frame out of every pile, and shows you exactly what it’s about to free before it frees anything.
Biggest win first, always. These are the numbers from a real six-year phone history — 48,000 photos, 214 GB, nothing unusual about it.
Nothing is double-counted. A blurred frame inside a burst is claimed by one pile only, so the total at the top is a number you can actually reach — not a sum of overlapping estimates.
Spotting the near-identical pile is the easy half. Keeper picks the frame you would have picked — and tells you why in four words, not a score out of a hundred.
Favourited, edited, cropped, filed in an album, shared. These outrank every algorithm in the app. If you touched a frame, it wins its group — and Keeper says that’s why.
Sharpness on the subject rather than the whole frame. Blown highlights and crushed shadows. Resolution and format. Faces in focus, eyes open, expression readable.
Most people present, because the shot with everyone in it is usually why the burst exists. Widest framing when the rest are crops. Latest in the burst — the one you waited for.
When it’s a close call, Keeper says so. Two frames neck and neck means neither gets pre-ticked and you get the final word — a confident wrong answer costs more than an honest shrug.
Every photo app promises this and you have no way to check any of them. Keeper’s version is built into the app, not written in a policy — and you can verify it yourself in about ninety seconds.
The app ships without permission to open an outgoing connection. There is no socket. The entitlement list is printed in Settings so you can check it against the shipped binary yourself.
No analytics SDK, no crash reporter, no remote config, no A/B framework — the four things that quietly turn a private app into a telemetry pipe.
Similarity, quality scoring, face detection and text recognition all run on the Neural Engine in your hand. No model downloads, no server-side inference.
Hashes and thumbnails live in the app’s own container, readable only while the device is unlocked. Never in iCloud, never in a shared folder. One button wipes it.
Keeper holds read access for its whole working life. Write access is used exactly once per session — the deletion you confirmed, through the system’s own dialog.
No sign-in, no sync, no subscription. All fourteen App Store privacy categories answered “Data Not Collected”, because there is nothing to collect it with.
Deleting the wrong photo isn’t like deleting the wrong file. So the safety net is the product, not the small print underneath it.
Two things people always ask a cleanup app for. Neither one can be done honestly as asked — so here’s what Keeper does instead of guessing.
In design, being built in the open. Each stage stands on its own, so there’s a real app in your hands at the end of the second one.
A command-line spike over a real library: group the duplicates, time every stage, check it holds up at 48,000 photos on phone-class silicon.
Similar sets, screen recordings, screenshots, exact duplicates, Recently Deleted. The confirm sheet, the receipt, undo. Mac with full keyboard culling.
Blurred and dark frames, long-tail video, forwarded media, the escrow export, and Apple Pencil strike-through on iPad.
Near-duplicate video, and learning from the times you overruled the pick — kept on device like everything else.
Every category, every screen, the scoring model, the on-device pipeline, and an honest list of what might not work. Plus the design principles that came out of drawing it.
Asking for your email would be a strange way to open a page about not collecting your data. Keeper turns up here and on the App Store when it’s ready — that’s the whole plan.
One Swift core shared across the three platforms, with the whole photo pipeline running on Apple’s own on-device frameworks.