FolioScribe

What FolioScribe does

FolioScribe is a web platform for transcribing handwritten archival documents. Upload a photograph or scan of a manuscript—a ledger page, a letter, a parish register, a company minute book—and FolioScribe returns a transcription together with confidence markers, uncertainty flags, and a side-by-side view of the original image.

The platform is designed around the working patterns of archival researchers: sustained engagement with difficult palaeographic material, iterative correction, and the need to cite original images alongside their transcriptions.

Who it is for

• Historians working on extensive archival corpora who need first-pass transcriptions they can refine.
• Archivists seeking to generate searchable finding aids for manuscript collections.
• Genealogists tracing family records across parish registers, census returns, and personal correspondence.

Technical notes

FolioScribe runs on a locally hosted vision-language model (Qwen2.5-VL), exposed to the public web through a Cloudflare Tunnel. The self-hosted architecture keeps uploaded documents on infrastructure under my direct control and avoids routing potentially sensitive archival material through third-party commercial services. Transcriptions include confidence flags ([?] for uncertain readings, [illegible] for unrecoverable text) that preserve the epistemic provisionality of palaeographic work.

Context

FolioScribe grows out of my own archival practice. Working with British, Belgian, and Portuguese corporate records for the monograph and accounting-history papers routinely produces more handwritten material than any single researcher can transcribe by hand. The platform generalises that working tool for others facing the same problem. It is a methodological contribution to the digital humanities rather than a commercial product.

Selected transcriptions produced using FolioScribe are made available under Primary Sources.

Citation

If FolioScribe contributes to your research, I would be grateful for acknowledgement. A suggested citation:

Primmer, A. FolioScribe: AI-assisted handwritten text recognition. University of Bristol, 2026. folioscribe.andrewprimmer.com.