Back to browse
GitHub Repository

Snap any image, screenshot, or webpage into plaintext. No GPU. No cloud. One command.

88 starsPython

CPU-only fast OCR for screenshots, images, PDFs, webpages

by mrkn1·May 31, 2026·9 points·8 comments

AI Analysis

●●SolidSolve My ProblemCozy

CPU-only VLM OCR beats Tesseract accuracy without sending data to the cloud.

Strengths
  • Heavily quantized 0.9B ONNX model runs on plain CPUs without CUDA dependencies.
  • Clipboard input and output workflow makes screenshot-to-text round trips incredibly fast.
  • Single-file portable design allows dropping the tool anywhere without installation requirements.
Weaknesses
  • Webpage processing isolates main content but could miss complex multi-image layouts.
  • 890 MB initial download is heavy compared to standard Tesseract installations.
Target Audience

Developers needing private, local OCR without GPU dependencies

Similar To

Umi-OCR · Tesseract · Google Cloud Vision

Similar Projects