Chemical Structure Image Search
Paste a screenshot of any chemical structure — from a paper, a notebook, a slide, or a patent — and instantly identify the compound and search 1.3 trillion molecules. Free, no login, no API key.
Why image-based structure search matters
Chemistry lives in images. A reviewer screenshots a structure from a PDF. A medicinal chemist photographs a whiteboard sketch. A student copies a structure from a textbook figure. Until recently, none of that was searchable. You had to redraw the molecule by hand or guess the SMILES — error-prone work that should have been automated a decade ago.
OCSR.ai solves this directly. Our optical chemical structure recognition engine achieves 93.2% accuracy on handwritten structures and near-perfect accuracy on printed structures from papers and patents. The recognized SMILES is then piped straight into our chemical search engine over 1.3 trillion molecules — so a paste-and-press-Enter takes you from blurry screenshot to a full property profile in under three seconds.
How it compares
| Capability | OCSR.ai | ChemDraw | DECIMER | MolScribe |
|---|---|---|---|---|
| Image → SMILES | ✓ Yes | Manual redraw | ✓ Yes | ✓ Yes |
| Handwritten structures | 93.2% accuracy | — | Limited | Limited |
| Paste from clipboard | ✓ Yes | — | Upload only | Upload only |
| Direct database search | 1.3T molecules | — | — | — |
| ADMET / pKa / toxicity | ✓ Built-in | Separate paid add-on | — | — |
| Price | Free | $1,910+/year | Free (research only) | Free (research only) |
| No login required | ✓ Yes | License required | Self-host | Self-host |
DECIMER and MolScribe are excellent open-source recognition models, but they only produce SMILES — you still have to plumb them into a database and a property prediction stack yourself. OCSR.ai bundles all of that in one place.
From image to insight in three seconds
- Paste or drop the image. Screenshot from a PDF, photo of a whiteboard, or any structure file — we accept PNG, JPEG, and most clipboard sources directly. No upload form, no signup wall.
- Get the SMILES instantly. Our OCR engine recognizes the structure and returns the canonical SMILES in roughly one second on a typical compound.
- Land on the full compound page. The recognized molecule is matched against our database. If it's a known compound (most are), you arrive on a page like aspirin or imatinib with the full IUPAC name, identifiers, and computed properties.
- Drill into ADMET, pKa, mutagenicity. Every compound page shows absorption, distribution, metabolism, excretion, toxicity, mutagenicity flags, and predicted pKa sites side-by-side. No tab-hopping between SwissADME, ProTox, and OPERA.
- Export a clean PDF report. One-click PDF with structure, identifiers, properties, and references — ready to attach to a regulatory submission, a thesis appendix, or an internal memo.
Who uses chemical image search
Medicinal chemists
Recognize structures from competitor patents, literature, and conference posters without manual redraw. Get ADMET context immediately.
PhD students
Digitize handwritten reactions from advisor whiteboards and notebook pages. Useful for thesis preparation and reproducibility.
Regulatory affairs
Quickly validate that submitted structures match known compounds. Generate clean PDF reports for filings.
Patent attorneys
Search prior art by structure rather than text. Catches near-equivalent compounds that wouldn't match a keyword search.
Frequently asked questions
Is this really free? What's the catch?
Yes — completely free for academic and individual research use, with no login or API key. OCSR.ai is built and operated by Rasayan Labs Inc., and the public site is part of our mission to make chemical intelligence accessible. We rate-limit bulk scraping but humans see no caps.
How accurate is the image recognition?
We benchmark at 93.2% on a handwritten-structure test set and substantially higher on printed structures from publications. Stereo, charges, and explicit hydrogens are handled. Complex polycyclics and unusual ring systems can still fail — when they do, our suggestion UI lets you correct the SMILES and re-search.
Can I batch-process many images?
Right now the web UI is one-at-a-time. We're building a Python SDK for batch workflows — email [email protected] if you'd like early access.
What about my data? Do you store the images?
Images are processed in-memory for recognition and aren't persisted. See our privacy policy for details.
Try it with a structure now
Open the homepage, paste any image of a chemical structure (Ctrl+V or Cmd+V), and watch the molecule resolve in seconds. No account, no API key, no rate-limit prompts.
Open OCSR.ai →