Upload a manuscript page and let AI transcribe it. Paleion combines specialized HTR engines with large language models to read historical handwriting that no one else can.
Drag and drop manuscript images in any format. Single pages or full collections.
Our pipeline detects the script type and selects the optimal recognition engine. LLM post-processing corrects errors.
Review the transcription with confidence scores. Export as plain text, TEI-XML, or other scholarly formats.
Supports Gothic, Humanistic, Procesal, Cortesana, and more. Automatic script detection selects the best engine.
Combines specialized HTR engines (Kraken, Transkribus) with LLM post-processing for maximum accuracy.
Trained on real archival documents from the 13th to 18th century. Not generic OCR — purpose-built for paleography.
Every transcription comes with word-level confidence scores so you know exactly where to focus your review.
Process entire collections at once. Upload hundreds of pages and get results without manual intervention.
Export as plain text, TEI-XML, or PAGE XML. Integrates with standard digital humanities workflows.
We're opening beta access in waves. Join the waitlist to secure your spot and get notified when it's your turn.
You're on the list! We'll be in touch soon.
I spent weeks trying to decipher a 16th-century procesal document. Paleion gave me a usable first draft in minutes.
The multi-engine approach is brilliant. Different scripts need different tools, and Paleion handles that automatically.
Finally a transcription tool that understands historical handwriting. The confidence scores help me focus my review time.
Paleion currently supports Gothic textura, Gothic cursiva, Humanistic, Procesal castellana, Cortesana, and several other historical script types from the 13th to 18th century. We are continuously adding new models.
Accuracy depends on the manuscript quality and script type. On well-preserved documents, we achieve Character Error Rates below 15%. Every transcription includes confidence scores so you can assess quality.
JPEG, PNG, and TIFF. We recommend high-resolution scans (300 DPI or higher) for best results.
Yes. Manuscript images are processed and deleted after transcription. We do not store your documents beyond processing time. All data is encrypted in transit.
Yes, on the Institution plan. You can provide ground truth transcriptions and we will fine-tune a dedicated model for your collection.
Paleion is a multi-engine platform. Instead of relying on a single HTR model, we combine multiple specialized engines and use LLM post-processing to correct errors. This hybrid approach consistently outperforms single-engine solutions.