# Dicklesworthstone/llm_aided_ocr

Enhances Tesseract OCR output using LLMs (local or API) for error correction, smart chunking, and markdown formatting of scanned PDFs

Repository: https://github.com/Dicklesworthstone/llm_aided_ocr
Canonical: https://ross.abutalabs.com/products/llm_aided_ocr
Language: Python
License: NOASSERTION
License Family: other
Topics: ai-assist, llama2, llm, ocr, tesseract, ocr-correction
Last push: 2026-08-03T04:19:28+00:00

## Health v2 (maintenance only)
Score: 71/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 95, release rhythm 35, longevity 81
- inputs: {"age_days": 1134, "days_push": 30, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2993, forks 214 (observed 2026-08-28T04:07:36.098840+00:00)

## What it is
A Python tool that converts scanned PDFs to text via Tesseract OCR, then uses LLMs (local or API-based like OpenAI/Anthropic) to correct OCR errors, remove duplicates, and format output as clean Markdown. It includes smart chunking, image preprocessing, quality assessment, and optional GPU-accelerated local inference.

## Use cases
- convert scanned pdfs to clean markdown
- fix OCR errors in digitized documents with an LLM
- correct tesseract output using local llama models
- turn old scanned letters or books into readable text
- remove headers and page numbers from OCR output
- batch process scanned pdfs into accurate text files

## When to choose
- you have scanned PDFs whose raw OCR output is error-ridden and needs cleanup
- you want Markdown-formatted output from scanned documents
- you prefer running correction with a local LLM for privacy or cost reasons

## When to avoid
- you need high-throughput OCR without LLM inference costs or latency
- you only need basic text extraction from clean digital PDFs
- you require a license-guaranteed dependency (license is non-standard)

## Facets
- artifact type: cli-tool
- maturity: active
- function: ocr, nlp, llm-inference, pdf, markdown, image-processing
- domain: pdf, developer-tools, large-language-models
- platform: python, cli
- tags: tesseract, ocr-correction, pdf-to-markdown, llm-post-processing, document-digitization, natural-language-processing, linux, macos, gpu

## Member repositories
- Dicklesworthstone/llm_aided_ocr (main) score 71

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:36.098840+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:47:26.433947+00:00, confidence not recorded.
  - readme: https://github.com/Dicklesworthstone/llm_aided_ocr (fetched 2026-08-28T04:07:36.098840+00:00, sha 4cffaf8fdd6c)
- Data as of 2026-08-30T08:39:29.467469+00:00.