Disclosure up front. We make PicDock, one of the apps below. So we are not neutral. In exchange, everything here is checkable — including the things our app does not do.
Nobody wants their whole camera roll on the living room wall. A photo library that goes back twenty years contains receipts, screenshots, a document you photographed at the bank, and eleven near-identical shots of the same plate of food. What people actually want on a frame is narrower: the family. The kids. The dog.
Apps have started offering exactly that — filter the slideshow down to certain people. It is a genuinely useful feature and it rests on face recognition, which means the interesting question is not does it work but where does it run. Faces are biometric data. Before you point an app at two decades of family photos, four things are worth knowing.
The four things
1. Does the analysis run on the device?
There are two designs. Either the app carries the recognition models and does the work on the tablet, or it uploads your photos to a server and gets answers back. Both can produce the same slideshow; only one keeps the faces of your children on a device in your house. The listing rarely says which — look for the words on-device or offline, and check whether the feature still works with the network off.
2. Does it group people automatically, or do you tag them?
Tagging a few hundred photos by hand is a project nobody finishes. Automatic grouping means the app clusters similar faces by itself and then simply asks you to name the groups. That is the difference between a feature you use and a feature you intend to set up one day.
3. Can you fix mistakes — and do the fixes survive?
Automatic grouping is never perfect. One person split into two groups, or two siblings merged into one, is normal. What matters is whether you can merge and split by hand — and then whether the app leaves your corrections alone. An app that silently re-groups everything the next time it analyses new photos will quietly erase an afternoon of tidying.
4. Which of your photos actually get analysed?
This one is easy to miss. If your photos live on a NAS or in a cloud account, an app that analyses only what is stored locally will find far fewer faces than you expect — and the filter will look broken rather than limited. Ask which sources the analysis covers before you judge the results.
How PicDock does it
About our app, precisely — this is a Premium feature:
- Everything runs on the device. The face detection and the recognition model are bundled into the app itself, so the work happens on your tablet and no photo and no face is uploaded anywhere for this. It works with the network off.
- Grouping is automatic. You are not asked to tag anyone; the app clusters similar faces and you name the groups you care about — or none of them, and just pick which ones to show.
- You can merge and split by hand, and we then stop re-grouping. Once you have edited the groups, PicDock does not run automatic re-clustering again — adding a few photos will not undo an afternoon of tidying. If you ask for a fresh grouping yourself, your names, your hidden people and your filter selection are carried across.
- Pets are a yes/no filter, not individual animals: PicDock can show only photos that contain a pet. It cannot yet tell your dog from the neighbour's.
- Videos are handled separately — faces found in videos are attached to existing groups rather than forming groups of their own, because a few sampled frames are weaker evidence than a photo.
Where it does not work
Two honest limits. Analysis covers photos that are already on the device — local albums, favourites, and the Google Photos selection the app has cached. Photos that live only on a NAS, in Dropbox or in OneDrive are not analysed unless they have been downloaded, so the people filter will not cover them. And the grouping was tuned on one library of about 490 photos on a Galaxy S20; it behaves well there and we have not verified every other shape of photo library. If it groups your photos badly, that is a real possibility, and the manual merge and split are there for exactly that reason.
In short
If you want a frame that shows only certain people, ask four questions: does the analysis run on the device, does it group automatically, do your corrections survive, and which photos does it even look at. The first is about your family's faces; the other three decide whether you will still be using the feature next month.
Related: Fotoo vs Frameo vs PicDock — a full comparison.
PicDock — A photo and video frame for Android. People and pet filters analyse faces on the device, and are part of Premium; the slideshow and all photo sources are free.