Structure OCR

Uploading images

Structure OCR lets you extract a chemical structure from an image. Click the image upload button in the toolbar (camera icon) and select a file.

Supported image formats: PNG, JPEG, and WebP. Maximum file size is 10 MB.

AI extraction process

When you upload an image, the AI’s vision model analyzes it and extracts a SMILES representation of the chemical structure. This SMILES is then validated through RDKit before being presented to you.

The result appears as a card showing the original image thumbnail alongside the extracted structure. The card includes:

• • RDKit validation status — "Structure validated (RDKit)" if the extracted SMILES parses correctly, or "Structure validation failed" if it does not.
• • Confidence level — Always "AI Suggested" with the message "Verify structure before use," because OCR extraction is inherently approximate.
• • Load to Canvas button — You must explicitly click this to load the extracted structure. OCR results never auto-load.
• • Try Again button — Re-run extraction if the first attempt was incorrect.

When to use OCR

Structure OCR is useful when you encounter a chemical structure in a paper, presentation, or textbook and want to work with it in ChemStitch. Upload the image, verify the extracted structure, load it to the canvas, and then use the AI tools to analyze it.

For best results, use clear images with high contrast between the structure and background. Hand-drawn structures and low-resolution images may produce less accurate extractions.

Rate limits

OCR is rate-limited to 10 requests per minute per user. This is sufficient for normal use — each extraction takes a few seconds, so the limit only affects rapid repeated submissions.