Your CZI file is 40 GB. Your LIF file is 80 GB. Your collaborator uses Nikon, you use Leica, the core facility uses Zeiss, and nobody can open anybody else’s files without the right software license. Your cloud analysis pipeline can’t stream a region without downloading the whole thing first.
This is not a niche problem. This is Tuesday in a microscopy lab.
OME-Zarr exists to fix all of this.
The old way: one giant blob
Traditional microscopy formats store data as monolithic files. TIFF, CZI, LIF, ND2, NDPI: one large file, all data inside. To display a single field of view at a specific zoom level, the software seeks through the file, decompresses blocks you don’t need, and hopes the metadata is where it expects it.
This was fine when datasets were small and everyone sat next to the same network drive. It is not fine when datasets are terabytes and your analysis runs in the cloud.
The deeper problem is fragmentation. The OME Bio-Formats library currently supports over 160 file formats. That number is not a success story. It is a monument to two decades of every microscope vendor inventing their own container format, their own metadata scheme, and their own compression quirks. Bio-Formats is the duct tape holding the whole thing together.
What OME-Zarr actually is
OME-Zarr (also called OME-NGFF, for Next-Generation File Format) flips the model completely.
Instead of one large file, data is stored as a directory of small, independently addressable chunks. Each chunk is a separate file: a 3D tile of pixels, compressed independently, stored flat. Want to display a 512x512 region at zoom level 3? Fetch the two or three chunks that cover that region. Nothing else.
This is exactly how Google Maps works. Different zoom levels are pre-rendered as tiles. You only load the tiles you can see. OME-Zarr applies the same idea to microscopy: every image ships with a built-in resolution pyramid, each level stored as its own set of chunks.
The result: partial loading, instant zoom transitions, and native compatibility with cloud object storage. S3, Azure Blob, Google Cloud Storage all work out of the box because they are designed for this exact pattern: small HTTP range requests against many small objects.
As the original Nature Methods paper puts it, “as latency of access grows, access times for monolithic formats such as TIFF increase because libraries must seek the appropriate data chunk, whereas NGFF formats such as Zarr provide direct access to individual chunks.” [1]
What the format stores
OME-Zarr is a metadata convention layered on top of Zarr, an open N-dimensional array storage library. The data is in Zarr: compressed, chunked, multi-dimensional. The OME part adds the science:
- Multiscale pyramid: every resolution level listed with its downsampling factor
- Physical coordinates: axes with names, types, and units (micrometers, seconds, channels); coordinate transforms that let you navigate in real physical space, not pixels
- Channel metadata: names, colors, contrast limits, all stored with the data
- Segmentation labels: masks stored in a
labels/group alongside the raw image; your segmentation and your data travel together - HCS layout: plate, well, and field-of-view hierarchy built into the spec; a 96-well plate is a single OME-Zarr store
One format. One metadata standard. No more emailing your Zeiss rep to ask what field 12 means.
Who is behind it
OME-Zarr was introduced in a 2021 Nature Methods paper by Josh Moore, Jason Swedlow, and colleagues at the University of Dundee and EMBL. [1] The Open Microscopy Environment (OME) consortium has been building open microscopy standards since 2005, when Goldberg, Swedlow, and colleagues published the OME data model. [5]
The 2023 community paper documents the rapid growth: dozens of institutions, hundreds of tools, terabytes of public data, and co-authors from EMBL, Janelia Research Campus, Chan Zuckerberg Biohub, Euro-BioImaging, and more. [2]
The Chan Zuckerberg Initiative funds active development. Jason Swedlow joined CZI in 2025. EMBL hosts public OME-Zarr datasets. The BioImage Archive at EMBL-EBI, one of the world’s largest repositories for biological imaging data, actively converts datasets to OME-Zarr for streaming access and visualization. [2]
This is not a scrappy side project. It has institutional backing, a governance process, and a roadmap to version 1.0.
The adoption numbers
- The 2024 OME-NGFF Challenge converted hundreds of terabytes of bioimaging data across multiple institutions in four months [3]
- The BRAIN Initiative (BICCN/BICAN) is building services to make thousands of whole-brain datasets available in OME-Zarr [2]
- The IDR (Image Data Resource) converts published datasets to OME-Zarr for browser-based access, publicly accessible over HTTP
- napari, Fiji/ImageJ, OMERO, webKnossos, and Vizarr all support it natively
- OpenOrganelle at Janelia uses it for FIB-SEM connectomics datasets at single-nanometer resolution [2]
The field has voted. It is not even particularly close.
What’s coming: Zarr v3 and OME-NGFF 1.0
Zarr v3 adds sharding: a way to store thousands of small chunks inside a single larger object. Previously, tiny chunks meant a massive number of files, which overwhelmed S3 and filesystems. Sharding decouples chunk count from object count. Fine-grained random access without the storage explosion.
Glencoe Software has published a roadmap to add Zarr v3 support in bioformats2raw and raw2ometiff. [4] OME-NGFF 1.0 is in active development via an RFC process: coordinate transforms for multimodal datasets, collections support, and a proper extensibility mechanism are the headline additions.
The format is not frozen. It is getting better.
See it in action
Open a real dataset right now. Find Nuclei Viewer streams any OME-Zarr file directly in the browser, no download, no install. Hit “Open Example” to pick from a curated list of IDR datasets, or paste any public OME-Zarr URL.
Your own data is still in CZI? The next post covers exactly how to convert it.
References
[1] Moore, J. et al. OME-NGFF: a next-generation file format for expanding bioimaging data-access strategies. Nature Methods 18, 1496–1498 (2021). https://www.nature.com/articles/s41592-021-01326-w
[2] Moore, J. et al. OME-Zarr: a cloud-optimized bioimaging file format with international community support. Histochemistry and Cell Biology 159, 223–251 (2023). https://pmc.ncbi.nlm.nih.gov/articles/PMC9980008/
[3] OME 2024 NGFF Challenge. https://github.com/ome/ome2024-ngff-challenge
[4] Glencoe Software. NGFF Roadmap 2026 (November 2025). https://www.glencoesoftware.com/blog/2025/11/06/ngff-roadmap-2026.html
[5] Goldberg, I.G. et al. The Open Microscopy Environment (OME) Data Model and XML file: open tools for informatics and quantitative analysis in biological imaging. Genome Biology (3 May 2005). https://pmc.ncbi.nlm.nih.gov/articles/PMC1175959/