EU inference
All models run through Cloudflare AI Gateway in front of EU providers. Zero data retention where supported.
01BioVision · Orasis
Every open-source microscopy tool. One terminal. Now with AI.
For thirty years, microscopy analysis has lived in a graveyard of incompatible plugins, dead Python environments, and macros that break the moment lighting changes.
Orasis — from the Ancient Greek ὄρᾰσις, "the act of seeing" — brings the best open-source engines under one roof: Cellpose, Fiji, scikit-image, ilastik, MyoFuse. Then puts an AI on top that actually understands what you're looking at.
Native Support
Drag-and-drop ingestion for imaging files up to 4 GB. Natively reads .czi, .tif, .ome.tiff, .lif, and .nd2 — no conversion, no plugins.
02The engines
A single terminal, the finest open-source engines.
Orasis unifies industry-leading algorithms under a single, cohesive interface, streamlining complex analytical workflows.
Fiji & scikit-image
Core processing and rigorous quantitative feature extraction (v2.14 & v0.24).
ilastik
Advanced pixel classification via interactive Machine Learning (v1.4).
Cellpose 3 & 4
State-of-the-art deep learning models for generalist cyto and nuclei segmentation.
Cellpose SAM
A robust foundation model capable of segmenting diverse biological images with unprecedented generalisation and minimal fine-tuning.
MyoFuse
A highly specialised plugin engineered specifically for the analysis of multinucleated myotubes and internal nuclei positioning.
03Adaptive AI · Agentic chat
From rigid macros to a lab partner that talks back.
Traditional image analysis relies on static macros — rigid scripts that break the moment lighting, magnification, or sample density shifts. Orasis still lets you save and replay macros, but with one decisive difference: the AI adapts them.
Hit "Adapt" and Orasis looks at the new image — contrast gradients, morphological variance, signal-to-noise — and re-tunes the pipeline on the fly. Same intent, new sample, accurate result. No more rerunning a macro on twenty fields just to discover the threshold drifted on field three.
And when you don't want to click at all, there's the chat. Type what you want. The AI sees the image, picks the right engine, adjusts parameters, executes, and shows its work. "Segment the nuclei, count them, flag anything that looks apoptotic." That's the command.
A vision-language model that actually helps — one that lives next to your data, reads the image, and has hands on the segmentation engines. Suggestions are sourced, parameters are visible, the final call is always yours.
The old way · static macros
The new way · just write
04Bench AI · Mobile
Your phone is now a microscopy analyst.
Not every analysis happens at a workstation. Half of microscopy lives at the bench — through an eyepiece, under a gel doc, holding a phone in one hand and a pipette in the other. So we built for that too.
Coming Soon · Mobile
Cytoscope for Bench
Point your phone through the eyepiece. Snap. The image is sent to our EU inference cluster, where Cellpose SAM does the heavy lifting, and the segmented result lands back on your screen in seconds — confluence, cell counts, the lot.
For cell-culture checks, passaging decisions, and the "is this confluent enough yet?" question you ask twelve times a day. No laptop required.
Coming Soon · Mobile
Gel Reader
Photograph a Western blot or DNA gel on your phone — angle, glare, fingerprints and all. The image is processed on our EU inference cluster: perspective corrected, lanes and bands identified, quantification returned with confidence intervals. Results land on your phone in seconds.
Built on the same lineage as GelGenie and the open-source gel-analysis community, with a careful AI layer on top. The reading is sourced. The disagreement is yours.
How we keep AI honest
Cautious by construction.
Frozen models
We don't fine-tune on your data. The same model the next user gets is the one you're using today.
Never the final word
Confidence is shown. Disagreement is logged. The note is yours, not the model's.
You can always disagree
One tap to flag a Biovision answer as wrong. We log it; the next reading benefits from your correction.
Standing on giants
None of this exists without open source.
Orasis doesn't replace the tools you already trust — it stitches them together and adds an AI that can drive them. Every engine below is the work of researchers and maintainers who have given their code to the community. Thank you.
Cellpose · Cellpose SAM
Generalist cell & nuclei segmentation. Stringer, Pachitariu et al.
Fiji / ImageJ
The bedrock of scientific image analysis. Schindelin, Rueden et al.
scikit-image
Quantitative image processing in Python. van der Walt et al.
ilastik
Interactive ML pixel classification. Berg, Kutra, Kroeger et al.
GelGenie
Deep-learning gel band quantification. The inspiration for Gel Reader.
MyoFuse
Multinucleated myotube and fusion-index analysis.
OME · Bio-Formats
Microscopy file format interoperability. Linkert, Rueden et al.
napari
Multi-dimensional image viewer that taught us a lot about UX.
Coming with the alpha
Join the waitlist for Biovision.
Biovision rolls out as part of the Aspis Bio alpha. Drop your email to be notified when your spot is ready.
You're on the list.
We'll email you when your spot opens up.