Digitize family travel journals and audit handwritten names
At a glance
Bryan Yeager converts photographed camping journals into structured records, then checks AI transcriptions against the originals. This workflow is listed under Family History, Digitization, OCR.
- Agents / tools
- Gemini 3 Flash, Python, VS Code
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Photograph every journal page in order and keep each journal's images in a separate folder. With Python and Gemini model access available, ask Gemini to help build a transcription script for those folders. 2) Process images sequentially with Gemini 3 Flash at Low visual reasoning, carrying context between pages. Request JSON containing the transcription, marginal notes, people, locations, page and journal numbers, and dates; iterate on the schema before processing the collection. 3) Generate a local fuzzy-matching script for names and an LLM proofreading pass to flag suspicious spellings. Compare proposed corrections with the photographed pages and family knowledge, then apply confirmed edits in VS Code. 4) Run a further scan for rare names to catch remaining inconsistencies and retain the corrected JSON journal collection.
Worth knowing
The article describes the script's logic but does not provide a complete runnable script. Human review remained necessary for ambiguous handwriting.