Finding photos again

Applies to: Photo Date Rescue for Windows, version 3.0.4. Last verified: 28 July 2026. Every statement on this page was checked against the shipping 3.0.4 build — both the source and the packaged bundles — and deliberately not against our own feature list. If you are on an older or newer version, the behavior may differ; check What’s new for changes.

This page is the reference we cite from our comparison table. Each capability below has its own permanent link, so a claim about PDR can always be traced to the paragraph that supports it. Where the honest answer is “no” or “only partly”, this page says so in the same words the comparison uses. We would rather publish a limitation about ourselves than a strength we cannot stand behind.

One thing to get straight before the detail. PDR is a repair and organizing tool first: its job is to date a pile of photos correctly and lay them out in plain folders. Search is what you use afterwards, on the library it has indexed. It is not trying to be your day-to-day photo browser, and this page does not pretend otherwise.

---

At a glance

---

Searching on the facts you can state

When you know something definite about the photo, PDR searches it directly out of its own index. All of the following are real filters, and they can all be applied at once:

• Filename and original filename, including partial matches.
• A date range, or a range of years or months.
• Camera make, camera model and lens model.
• Media type and file extension.
• Country, city, or simply “has GPS” or “has no GPS”.
• ISO, aperture, focal length, megapixels and file size, each as a from-and-to range.
• Orientation and scene capture type.
• Albums you have made.
• The AI tags PDR generated, if you have switched that on.

Three limits on this, stated plainly:

• There is no star rating in PDR — not in the database, not in the interface. So there is nothing to search. If you sort by picks and rejects today, that habit does not survive the move.
• You cannot search by subfolder. The only location filter is the whole indexed library drive. There is no “only inside this folder” search.
• Caption text is not searchable. You can filter for photos that have a caption, but the words you typed into it are not indexed, so searching for them finds nothing. This is a genuine gap and it has been open for several versions.

---

When you only half-remember it

If you cannot recall a filename or a date, PDR gives you one route and withholds two others.

What works. With the AI features switched on, PDR looks at the content of each picture and tags it. Typing a word like sunset, beach, wedding or dog will then find photos on what is actually in them, not just on their filenames. Free text also searches the names of people you have identified.

The limit that matters. That tagging works from a fixed built-in list of 193 terms. It is not open-ended. A word outside the list finds nothing at all — so “sunset” works and “the evening we walked back along the harbour” does not. If you want to describe a scene in your own words, PDR is not the tool for that today.

Two things PDR does not do at all:

• It does not read text inside pictures. There is no way to find a receipt by the shop’s name, a sign by its wording, or a document by a line in it.
• It does not search video speech. PDR can transcribe what is said in a video, and will show you that transcript, but those words are not connected to the search box — so you cannot find a clip by something said in it. The transcript is stored and displayed, and that is where it stops.

---

Stacking conditions, and keeping the query

Conditions combine properly. Different filters are joined with and, so adding one always narrows the result. Picking several values inside a single filter is an or. The search box understands a small grammar on top of that:

• Terry + Mel (or &, or the word and) — photos with both.
• Terry, beach — photos with either.

As you build a query, PDR shows you how many photos each further option would leave, so you can see where a condition is about to take you before you click it.

Keeping it. A whole set of filters can be saved as a named favorite and reopened later, and favorites can be renamed and deleted.

The limit. There is no general way to exclude things. There is no “not this person”, “not this tag” or “not this word”. A handful of fixed opposites exist — no location, no camera, no lens, no faces, flash did not fire, not yet processed — but that is the whole of it. If your way of narrowing a big result is to subtract, PDR will frustrate you.

---

Finding a person

PDR detects faces and groups them. Name a face once and it will pull up every photo of that person, including ones it matched automatically. Because the confidence of an automatic match varies, you get two controls: a switch to show only faces you have confirmed yourself, and a slider to tighten or loosen how similar a face has to be before it counts.

Once you have a person, the rest of the search still applies, which is the point of this section. You can narrow the same query by date range, by city or country, by album, or by an AI tag such as wedding. You can also ask for photos containing several named people together, or photos of just one person on their own.

Pets are not included. PDR’s face detection is for human faces only. A dog can be found as dog from the general tag list, but never as Bella. If naming your animals matters to you, other tools do this and PDR does not.

---

Keeping the index honest

PDR’s index is a real database file on your machine, so it survives closing the app and restarting your computer. It also tries to tell you when it has lost sight of something:

• A library drive that is not connected is labeled Drive not connected, and a folder that has moved is labeled Folder no longer exists.
• When a previous run’s location has gone, PDR offers to point it at the new location or remove it.
• If it finds photos on disk that are not in the index, it says how many and offers to index them.
• After a Fix run, new results are indexed automatically by default.

Three limits, and the last one is the important one:

• Search results themselves carry no warning. Those labels describe indexed locations, in the Search Index Manager. They are not attached to individual results. If a photo’s drive is unplugged while you are searching, its result still appears, still shows a thumbnail from a cache, and asking to open its folder does nothing at all — with no message to tell you why. Corrected 30 July 2026: this section used to imply the labeling reached your results.
• PDR does not watch your folders live. Nothing is monitored in the background. The index refreshes after a Fix run, or when you ask it to.
• If PDR starts up while a library drive is unplugged, it removes those photos from the index rather than holding them as temporarily offline. Nothing on your drive is touched — the photos themselves are perfectly safe and untouched. But re-indexing is not the whole cost, and we previously described it as though it were. Removing those entries also removes the faces PDR had recognized in them, the subject tags, any captions you had typed, their album memberships and their video transcripts. All of that has to be built again, and the analysis is the slow part. Corrected 30 July 2026. If you keep your library on a drive you unplug, open PDR with the drive connected.

---

Where the searching happens, and what leaves your machine

All of the searching happens on your own computer. The index is a database file on your disk. Face recognition, content tagging and video transcription all run on your own hardware. Your photos are never uploaded, and no service anywhere receives your images or your index.

The AI features are switched off until you turn them on. When you do, PDR downloads the recognition models once — that is the only time it needs the internet for this — and after that it works without a connection.

For completeness, here is every reason PDR contacts the internet at all. None of them involve your photos:

• Downloading the AI models, once, when you first enable them.
• Checking your license with our payment provider.
• Checking for updates, and counting files against the free trial limit.

PDR collects no analytics and no telemetry of any kind. There is no PDR account: no sign-up, no email and password, no profile. There is also no server edition — PDR is a desktop application, so “running it on your own machine” is the only way it runs.

---

Where you can search from

By keyboard, on the desktop, it is well covered. Ctrl+F jumps straight to the search box, the arrow keys step through your results, Enter opens the selected photo and Escape backs out. The dropdown filters and the person picker are navigable with the arrow keys and Enter too.

By device, it is narrow, and this is the honest headline: PDR is Windows only. There is no Mac build, no Linux build, no browser version and no phone app. If your photos need to be searchable from the sofa or from your pocket, PDR does not do that and several products in our comparison do.

One thing that sounds like it might help but does not: the “send to phone” feature hands over photos you have already selected on the PC, over your local network, for fifteen minutes. It is a hand-off, not a phone client. You cannot search from the phone.

---

Running the recognition work on your own machine

Yes. Every part of finding a photo happens on your computer. Searching, filtering, face recognition, recognizing what is in a picture and transcribing the speech in a video all run locally. No query and no image is sent anywhere to be processed, and there is no service that sees what you search for. The section on where the searching happens lists every reason PDR contacts the internet at all; this one is about what running locally costs you in practice, because it is not free.

To do it

1. Switch the AI features on. They are off until you do, and nothing is recognized until they are.
2. Let the first download finish. A recognition feature needs a model on the machine before it can work, and the first time you use one it fetches it — a substantial download, and one you need a connection for. It is a download to you, not an upload of your photos.
3. Search normally after that. The recognition itself never needs a connection again.

Limitations

• The first use of a recognition feature is not an offline action. The model has to arrive before anything can be recognized, and the files are large. On a slow or metered connection this is a real wait and a real cost, and there is no way to get the recognition features without it.
• Recognition runs on your processor, not your graphics card. There is no option in this version to hand the work to the graphics card. The result is thorough but slow: on a large collection, expect a long first pass rather than a quick one.
• Saved searches stay on the machine that made them. Nothing you build is synchronized anywhere, which is the same fact seen from the other side — see keeping a search.

---

Deciding how sure a face match has to be

Yes. A sensitivity control sets how confident a match has to be before PDR shows it as that person. Slide it towards strict and you see only strong matches; slide it loose and you see more, including some that are wrong. The results update as you move it, and nothing is reprocessed while you do — you are changing the threshold on work already done. Faces you have confirmed yourself are always shown, wherever the control sits, so tightening it can never hide a person you have named by hand. The section on finding a person covers what else you can narrow by once you have one.

To do it

1. Pick a named person, so there is a set of matches to judge.
2. Move the sensitivity slider towards strict until the wrong faces disappear. The results follow the slider immediately.
3. If you want certainty rather than coverage, turn on the switch that shows only faces you have confirmed yourself.

Limitations

• It changes what you are shown, not how faces were grouped. The grouping happened when the faces were processed, using a fixed setting you cannot reach or adjust. If two people were merged into one group, no position of this slider separates them.
• It is one setting for everybody. You cannot be strict about one person and relaxed about another. A face that is hard to match drags the whole library loose, or stays missing.
• It shows no number. There is no percentage, no score and no readout, so the control is a feel rather than a measurement. You cannot write down where you set it, and you cannot compare two settings except by looking at the results.

---

Seeing what a filter will do before you commit to it

Only partly, and the honest version is blunter than it sounds. Filters take effect as you set them and the count of matching photos updates live, so you can watch a filter narrow things down and back it off if it has gone too far. There is no separate Apply step to commit to, and nothing has to be confirmed. But that is the reason there is no preview: the count changes because the filter has already changed, not because PDR is showing you what would happen. The one genuine look-ahead is the number beside each option in a filter list, which tells you what a single next condition would leave — see stacking conditions.

Corrected 30 July 2026, twice and both times against us. Those look-ahead numbers appear on the list-style filters — camera, lens, file type, place, album, library drive and the shooting-metadata lists. Date ranges, the numeric ranges such as ISO or file size, people and subject tags have no number beside them, and we had written as though every filter carried one. The numbers also ignore any person or subject-tag filter you have already set, so with one of those active they read higher than the result you will actually get.

To do it

1. Set a filter and read the result count. It is the live count of what you now have, not a projection.
2. If the result is too narrow, take that filter off again. Undoing is how you explore.
3. To judge one more condition before applying it, read the number beside each option in the filter list rather than clicking through them — on the filters that have one, and provided you have no person or subject-tag filter set.

Limitations

• There is no “what would this do” preview. Every filter is applied the moment you set it. If you want to know the effect of a combination, the only way to find out is to apply it and look.
• On a very large collection the grid lags while it catches up. The count and the thumbnails do not arrive together, so a fast series of changes can leave you looking at results that are a moment behind the filters you set.
• There is no single action to clear everything. Filters come off the way they went on, one at a time, which is tedious after a long session of narrowing.

---

Keeping a search and opening it again later

Yes. A search you have built can be named and kept, and reopening it runs it again against the library as it stands now. That distinction matters: a saved search is a question, not a frozen list, so photos added since you saved it appear in the results the next time you open it, and photos that no longer match drop out. Favorites can be renamed and deleted — see stacking conditions for the query grammar that goes into one.

To do it

1. Build the search you want — filters, people, tags, a date range, or any combination.
2. Save it as a favorite and give it a name you will recognize in six months.
3. Reopen it whenever you want. It runs again from scratch, so the results reflect today’s library rather than the day you saved it.

Limitations

• Nothing tells you when new photos have started matching. There is no badge, no count and no notification on a saved search. It is re-run only when you open it, so you find out by looking.
• Saved searches stay on that machine. They cannot be exported, backed up on their own, or handed to somebody else. Move to a new computer and you build them again.
• They cannot be built on top of one another. There is no way to open a saved search and narrow it into a second saved search that stays linked to the first. Each one is independent, and a change you want in all of them is a change you make in each of them.

---

Known gaps in 3.0.4

Collected in one place, so you do not have to read the whole page to find the bad news:

1. No star ratings anywhere in the product, so none to search.
2. No subfolder search — the narrowest location filter is a whole library drive.
3. Caption words are not searchable, only their presence.
4. Video transcripts are not searchable. PDR makes them and shows them, but they are not wired to the search box.
5. No reading of text inside pictures.
6. No free-wording scene search — scene matching uses a fixed list of 193 terms.
7. No general exclusion in queries.
8. No named pets.
9. Starting up with a library drive unplugged drops those photos from the index and they must be re-indexed when it returns.
10. Windows only — no Mac, Linux, browser or phone.

---

How to check this page is still true

This page is cited as evidence in a public comparison against other people’s products, so it carries a higher burden than ordinary help text. Everything above was verified against the shipping 3.0.4 build on 28 July 2026 — reading the source, then confirming each claim survived into the packaged bundles, because a feature that exists in the source but cannot be reached in the shipped app is not a feature. Where that check found something unreachable, it is reported above as absent rather than described as working.

If something here does not match what the software does on your machine, that is a fault on our side and we want to hear about it. Please contact support. Tell us the anchor — for example /help/discover#vague — and we will either correct the software or correct the page.