What kind of descriptions work best?

Plain ones. Say what you’d see in the picture — who’s in it, where it is, what’s happening: “the kids on the beach at sunset”, “our table at the restaurant”, “the wide shot of the valley”. You don’t need tags, filenames or dates.

Can I search by date?

Yes — “photos from last summer”, “anything from 2023”, or “more from that same day” after a result all work, on their own or combined with a description.

Can I search for a specific person?

Yes, once you’ve tagged their face. Tag someone once and you can ask for “photos of Maya” from then on. Untagged, describe them instead.

Why do some results say “new”?

Those photos are in your album’s uploads but not on any page yet. They’re marked so you can spot the ones you can still add at a glance.

Does it change my album automatically?

No. Searching only shows you photos. Nothing moves until you apply a photo to a spot or ask for them to be placed — and you can undo it afterwards.

Sustain

← Back to Learn

What’s New

August 10, 2026

· 4 min read

Find any photo by describing it — no scrolling required

The Sustain album assistant answering “find photos with mountain views” with a row of photo thumbnails, one marked “new”

Every album has that one photo you know you took. The problem is finding it — somewhere in a few hundred thumbnails, in a month you can’t quite remember.

Now you can just say what it is. Ask the AI Album Designer for “the photos of the kids on the beach at sunset” and it comes back with the actual pictures — not a list of numbers to go and look up.

Describe it the way you’d say it out loud

Open the assistant in your album and type what you’re looking for. You don’t need keywords or tags — a plain description works, in whatever language you write in:

“find photos with mountain views”

“the ones where everyone is around the dinner table”

“photos of Omri from last summer”

“something for the cover — the widest landscape shot”

It works because every photo you upload gets described automatically when it’s analyzed. That’s what the search reads — so it can find “people around a table” even if nothing in the filename or the date says so.

You get pictures back, not numbers

The answer arrives as the photos themselves. A handful of matches come as cards you can read at a glance — the picture, a line about what’s in it, and a button to use it. A bigger set becomes a swipeable strip you flick through, so a broad search doesn’t take over the conversation.

Anything not in your album yet is outlined and marked new — usually the most useful thing on screen, because those are the photos you can still add.

Narrow it down by just saying so

The assistant keeps the thread, so you can whittle a result down the way you would with a person:

“only the ones with people in them”

“just the best three”

“what about the same day, but outdoors?”

Each reply re-searches with your refinement in mind, instead of making you restate the whole request.

Then put them in the album

Once you can see the photos, there are two ways to use them:

Swap one in. Select a photo on the page first, then hit Apply on the one you like — it takes that spot.

Ask for them to be added. “Add #2 to a new page” or “put these three on their own spread” and the assistant places them, choosing a layout that fits.

Nothing changes until you say so, and every edit can be undone from the chat.

Tip

Searching for a person? Tag their face once (in the People panel or via the assistant’s “Tag faces” pill) and you can ask for them by name from then on.

Good to know

It searches this album’s photos. Everything you uploaded to the album is fair game, whether or not it made it onto a page.

Newly uploaded photos need a moment. A photo is searchable once it has been analyzed — usually within a minute of uploading.

Names only work for tagged faces. Until you tag someone, ask by what’s in the picture instead (“the boy in the red shirt”).

It costs a few credits per message, like any other request to the assistant.

Curious how it works?

Behind that one line you type, a second AI agent goes off and does the digging — searching, reading the descriptions of the candidates, and throwing out the near-misses before you ever see a result. We wrote up that design in Nested AI agents, if you like knowing what’s under the hood.