Find the photo you remember.
A dog on the beach. A receipt from last year. Someone's birthday in Banff. Search the scene, the words in a picture, or the people and places you know, then narrow the results without sorting every photo by hand.

Ask in plain words.
Ask Search turns everyday phrases into filters: "photos of Alice in Banff last summer", "receipts from 2024", "screenshots from last month", "favorite videos since 2020". It understands dates, months, relative ranges, places, favourites, media type and the people you've named, and shows you exactly what it searched for.
Press ⌘K or / from anywhere. Results and a preview appear while you type, with a histogram of when, counts for every filter, and your recent and saved searches.

Ask Frameleaf Assistant.
Type a whole question and Frameleaf Assistant answers with the photos, and a line under each one saying why it matched: who is in it, where and when it was taken, words in the picture or something said in a video. It runs on your own server and never sees your Locked photos.
Prefer Claude or ChatGPT? Connect the assistant you already use and ask about your library from there.

Narrow it down.
Keep a useful search as a saved search, collect today's results in an album snapshot, or turn supported filters into a smart album that adds new matches. A snapshot shows its count before you create it; visual search saves its top ranked matches.
- People, places and dates with any-of, all-of and exclude groups.
- Text in photos on signs, receipts, whiteboards and screenshots, with the matching words highlighted.
- Description, tags and rating including descriptions written by AI.
- File name, folder, camera and lens when you remember the technical side.
- Favourites, archive and not-in-any-album for tidying up.

Find a photo from your phone.
The apps can search the catalogue already on your phone, then ask your server for smart search results when connected. Narrow by the filters available in your app. Offline catalogue results do not mean the full photo or video has been downloaded.


Albums that keep up with you.
Build a smart album from supported people, tags, dates, media types and visual searches. Preview its matches and choose whether additions need your review. For filters a rule cannot reproduce, keep a saved search instead.
Administrators can also enable built-in albums for Travel, Documents & Receipts, Screenshots, Food, Pets and Nature. They use available descriptions and search data. Remove a mistaken match from one of these albums and later updates keep it out.

Less labelling. More finding.
Smart search and text recognition use the processing your administrator configures. Optional AI descriptions add another way to find a photo; generated details are suggestions you can review and correct.
Descriptions
A plain-sentence description and searchable tags for every photo, added beside anything you wrote yourself, never over it.
Familiar names
Descriptions can use the names you've confirmed in People. Review face matches and generated text when a name is wrong.
Your vocabulary
Choose terse, balanced or rich, tell it what to look for, and give it your own words and instructions.
Video moments
Find matching moments in processed videos and jump to them. Sampled frames and available captions help you look inside a clip.
Find similar
Start with a photo you already found and look for other pictures like it, then add people or date filters.
Sensitive content
Optional detection marks likely sensitive photos and can hide them behind your PIN, with your review.
Your processing choice
Use compatible hardware on your server or another configured worker. Model choice and available hardware affect speed and results.
Your language
Multilingual models can search in more than one language. Language coverage depends on the model your administrator selects.
Documents, found.
Find the receipt with a shop's name on it, copy text from a sign, or browse recognised documents together. Text recognition must finish first, and you can correct words it misread.

Private by design.
Your server stores and searches your library's index. Machine learning can run on the same machine or a configured remote worker, which receives the images or search text needed for the job. An administrator can also choose Frameleaf Cloud to prepare smart search for servers without a capable GPU: it reads a metadata-stripped preview, and your server keeps the result. What you type in the search box is always read on your server.
Optional Cloud GPU processing has its own setup and consent. Locked controls what appears in your app session; it does not prevent server jobs or configured processing from reading those files.
