All your photos. Searchable. Seen by nobody.
Gallery is a real photo library for a family, a team or a whole archive: a grid that scrolls through a decade without stuttering, albums, a trash that gives you thirty days, and search that finds "beach" or "dog" without anybody tagging anything. The part that looks at your photos runs on your own device.
Nemi Gallery
Photos are made smaller on your device, then join the library.
Nemi Gallery, drawn from the app rather than filmed, so it stays true as the product moves.
What it does.
A private photo library you can search by what is in the picture.
Search finds the cat, and nobody else sees it
The model that recognises what is in a photo runs in your own browser, while the photo is being compressed. What we hold is about five hundred numbers per photo. No photograph of yours is ever sent anywhere to be analysed, and there is no facial recognition anywhere in Nemi.
Far more photos in the same storage
Photos are re-encoded to AVIF on the device that adds them, so what travels and what is stored is a fraction of the original size. (A browser that cannot do it sends the photo once and our server does the same job.) Gallery keeps that version and not the original file, which is what makes a very large library affordable.
One library per workspace
Everybody in the workspace sees the same grid, the same albums and the same trash. Your name is on what you added, and erasing something for good stays with you and the workspace owner.
Still fast at a hundred thousand photos
The grid only ever holds the two screens you are looking at, and thumbnails come straight from storage instead of through our server. Scrolling back ten years feels the same as scrolling back ten days.
Inside Gallery.
Every row ends on the same straight line
Photos are fitted so each row finishes flush with the edge of the grid, and nothing is cropped to get there: a panorama takes the room of three portraits and a tall photo stays tall. The last row of a month keeps its height rather than stretching, because one leftover photo blown up to full width reads as a mistake.













Type cat and the cats come back
Nothing has to be tagged first, and no photograph of yours is sent anywhere to be looked at. The part that reads a picture runs in the tab that is already compressing it, and what reaches us is 512 small numbers, about half a kilobyte. What you type is turned into numbers on your device too, so results arrive in order of likeness, which is why the grid marks where the confident matches stop.









The original is never stored
Each photo is re-encoded to AVIF in the browser that is adding it, at your plan's resolution, so what travels is a fraction of the original and the original itself is never stored. That is what makes a very large library affordable. Nothing the device claims is taken on trust either: once the file is in storage we check the stored file itself for the format and the size. A browser that cannot encode AVIF falls back to compressing on our side, which is the exception the rest of this is built to avoid.
Adding 412 of 800 photos
Compressing 6 on this device
It comes with
the rest of Nemi.
One login, one workspace, one bill. Every app below is included in the same account, on every plan.
All your work. One account.
Nemi is opening in groups. Put your name down once and the whole of Nemi is yours when it is your turn.
15 GB free when you are in. No card.













