ArchiveRescue — Restore Old Documents with Open-Source AI

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

I built ArchiveRescue, a tool that helps restore and digitize old scanned documents using open-source AI.

Users can upload documents, improve image quality, extract text using OCR, and download the results.

Features include:

• Image enhancement and noise reduction.
• OCR text extraction.
• Selecting specific areas for OCR.
• Editable transcription.
• Confidence warnings for inaccurate results.
• Exporting documents and text as ZIP files.

This could be useful for historians, researchers, museums, and freelancers offering document digitization services.

Demo

Live Demo: https://archiverescue.e1ite.workers.dev/

Upload a document, choose an enhancement mode, and click Restore & transcribe.

You can review the extracted text, make corrections, and download the results.

Code

GitHub: https://github.com/shadowe1ite/ArchiveRescue

The project is open-source under the MIT license.

How I Built It

I used several open-source tools:

• Tesseract.js — OCR directly in the browser.
• OpenCV — Image enhancement and noise reduction in the Python version.
• TrOCR — Experimental handwriting recognition using an open-weight model in the local version.
• Gradio — Interface for the Python version.
• Cloudflare Workers — Hosting the browser demo.

One challenge was recognizing handwritten documents. The initial OCR results weren't accurate, so I added region selection, image enlargement, and confidence warnings.

Handwriting recognition still needs improvement.

Why Does Open Innovation Matter?

I chose open-source AI because it allows documents to be processed locally without uploading them to third-party OCR services.

It also keeps costs low since there are no per-document API charges.

Using open-source tools gives me the freedom to improve the OCR pipeline, experiment with different models, and add new features without depending on proprietary services.

My Agent Session

I used AI coding assistance to help build and improve the project. I haven't recorded a DevRelay session.

My First User

I tested the application with a historical document. The OCR struggled with handwritten text, which helped me identify areas to improve.

I haven't had a paying user yet.

Prize Categories

Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.

---

Demo: https://archiverescue.e1ite.workers.dev/

Source Code: https://github.com/shadowe1ite/ArchiveRescue

Thanks for checking it out!