Local Files

Open OME-ZARR data from your local disk without uploading anything.

Reading from disk is the fastest and most private way to work, but it is not the only one. Your OME-ZARR Storage Setup Is Slowing You Down compares this with streaming from a server, and is worth reading before a lab or facility commits to one setup.

Which Browsers Can Do This

It depends on the shape your data is in.

What you open	Chrome, Edge	Firefox, Safari, iOS
A dataset folder	Yes	No
A dataset zipped into one file	Yes	Yes

Folders need Chrome or Edge

Reading a folder from disk needs the File System Access API, which ships in Chrome and Edge and nowhere else. Both folder routes below depend on it: the folder picker and drag and drop alike.

Firefox and Safari cannot open dataset folders. Neither can any browser on iOS or iPadOS, because they all run on WebKit.

A single file opens anywhere

A dataset zipped into one .ozx or .zip file is just a file, and every browser can read a file. Drop it in Safari or Firefox and it opens. See A Dataset in a Single File below.

The third route is to serve your data over HTTP and open it by address. See Remote URLs, including the free Data Server that puts a folder on your own machine behind a local URL in one command.

Drag and Drop

The simplest way to open local data is drag and drop, and it works in two places.

On the start screen. Open Find Nuclei Viewer in Chrome or Edge, drag your .ome.zarr folder from Finder or Explorer onto the browser, and drop it. The viewer opens it immediately.

On the viewer, while an image is open. Drop the folder straight onto the image. There is no need to close what you have first. A drag can go astray in a way a click cannot, so this drop asks before it swaps: the prompt names the folder you dropped and the image it would replace, and Cancel leaves your view exactly as it was.

Either way, a folder that holds datasets rather than being one is welcome too. The viewer offers you the datasets it found so you can pick one. See If the Folder Holds Datasets below.

Using the File Picker

Or use the built-in folder picker:

1. Click Open Image on the landing page
2. Click Open Folder in the Load Data modal
3. Select your .ome.zarr folder
4. The viewer loads your data

A Dataset in a Single File

An OME-Zarr dataset is thousands of small files in nested folders. That shape is what makes it fast to read one region without reading the whole image. It is also what makes it awkward to send someone or carry on a stick.

The format has an answer: pack the whole dataset into one ZIP archive. Name it .ozx, or leave it as .zip. Both open here.

To open one, drop it onto the viewer, or click Open Image and then Choose Zipped Dataset. It opens the way a folder does, with the same plate navigation, the same channels and the same speed. The viewer reads pieces of the archive as it needs them, so a large dataset never has to be unpacked first.

This route works in every browser, Safari and Firefox included. It is the only way to open data from your own disk in those two.

Making one

The quickest way is the one your computer already has. Right-click the dataset folder and choose Compress on macOS, or Send to › Compressed (zipped) folder on Windows. The zip holds the folder itself, and it opens here as it is.

For the smallest and fastest archive, zip the contents of the folder from the command line instead:

cd your-dataset.ome.zarr
zip -r -0 ../your-dataset.ozx . -x '.DS_Store' '*/.DS_Store'

That is the whole recipe. Three details in it matter.

. rather than the folder name puts zarr.json at the top of the archive, which is the .ozx layout other tools expect. The viewer opens it either way.

-0 stores the files instead of compressing them. Zarr chunks are already compressed, so compressing them again costs time on both ends and saves almost nothing. Stored entries are also the fastest to read a piece of.

The -x at the end drops the hidden files macOS leaves in any folder you have opened in Finder. They are harmless, and without it a plate picks up one in every row. Leave it off on Linux and Windows.

Any tool works. Name the result .ozx or .zip, whichever you prefer.

Check it opened the way you meant

Drop the archive on the viewer. If it opens, it is right. If you see a message that it holds no dataset the viewer can find, the dataset is neither at the top of the archive nor in a single folder there. Two datasets zipped side by side is the usual cause.

Double-click to open

Install the viewer in Chrome or Edge and it registers .ozx, so double-clicking one in Finder or Explorer opens it straight in the viewer. No browser window to find first, no dragging.

Only .ozx is registered, never .zip. A microscopy viewer claiming every archive on your machine would be presumptuous, and most zips you double-click are not datasets. Rename a .zip to .ozx if you want this for it.

If the viewer is already open, the file opens in that window and asks before it replaces whatever you were looking at.

To install: open the viewer in Chrome or Edge and use Install in the address bar. See Installing and Offline for what else that gives you.

Nothing is unpacked, and nothing is copied

Opening an archive does not read it. The viewer reads the ZIP index, then only the pieces it needs, seeking straight to them inside the file.

Opening the 96-well plate used for testing this, 10 MB with 205 entries, reads 0.08 MB to open it, 0.10 MB to work out the whole plate layout, and 0.41 MB by the time the first field is on screen. That is 3.9% of the file. The rest is never touched until you navigate to it. A dataset far too large to fit in memory opens exactly as fast, because size was never the thing being read.

Your data stays on your disk. It is not uploaded, and it is not copied into the browser either.

Why Safari and Firefox behave differently

Reading a folder needs the File System Access API. Chrome and Edge have it. Firefox and Safari do not, and there is no way around that from a web page.

A single file needs nothing special. Every browser can read one. That is why a zipped dataset opens in Safari and a folder does not.

The difference shows up again when you come back later.

	Chrome, Edge	Firefox, Safari
Open a zipped dataset	Yes	Yes
Reopen it from Recent, after a reload	One permission prompt	Pick the file again

Chrome and Edge can hand the viewer a lasting reference to your file, so returning to it asks only for permission. Firefox and Safari have no such reference to give, so the viewer asks for the file.

Either way you land back on the same image record, with your annotations, channel settings and contrast exactly as you left them. Only the way you point at the file differs.

Why we do not just remember the file itself

A browser can store a whole file in its own storage, and then Recent would work everywhere with no prompt. We measured what that costs: storing a 10 MB archive that way grew browser storage by 10.5 MB in Firefox. It copies.

For a plate of any real size that is a second copy of your data, written somewhere you did not choose and cannot easily see. Asking you to point at the file again is a smaller price, so that is what the viewer does.

If the Folder Holds Datasets

Data rarely sits on disk as one lonely dataset. Open a folder that holds several and the viewer says so, lists what it found, and opens whichever one you pick:

1. Drop the folder, or choose it in the picker, as usual
2. The viewer reads each folder inside it to see which are datasets
3. Pick one from the list and it opens

The names in that list come from the data, not from the folder names. A dataset called plate-3 or run-2024-05 is found just as readily as one called sample.ome.zarr.

Nothing is uploaded to work this out. The viewer reads a few hundred bytes of metadata from each folder, on your machine, exactly as it reads the dataset itself.

Privacy

Your data stays entirely on your machine. Find Nuclei Viewer processes everything locally in the browser. No files are uploaded to any server.

Recent Images

After opening a dataset, it appears in your Recently Opened list on the landing page. Click any recent image to reopen it. Thumbnails and metadata are cached in your browser's local storage.

Reopening asks you to confirm access, because the viewer holds a reference to your data rather than a copy of it. A folder and a zipped dataset in Chrome or Edge ask for permission. A zipped dataset in Firefox or Safari asks you to point at the file again. See Why Safari and Firefox behave differently.

Whichever it asks, you return to the same image record: annotations, channel settings and contrast are as you left them.

Related guides

• Remote URLs: the third route, and the one to use when the data lives on a server
• Data Server: serve a folder from your own machine over HTTP, so local data opens anywhere