skills.rest
AnderHonorato
AnderHonoratoCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

read-special-images

Extracts text and content from oversized or text-dense images via deterministic overlapping tiles.

Updated Aug 28, 2026
Search Tags:#ocr#text extraction#vision#pillow#image tiling#screenshot reading#image slicing
One-click install
npx skills add https://github.com/AnderHonorato/Mem-rias-IA---Infinity --skill read-special-images-anderhonorato
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: read-special-images
Source: https://github.com/AnderHonorato/Mem-rias-IA---Infinity/tree/main/Manus/Skills/read-special-images
Command: npx skills add https://github.com/AnderHonorato/Mem-rias-IA---Infinity --skill read-special-images-anderhonorato

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires Pillow, and includes scripts (resource) components.

What problem does it solve? Whole-image previews of very tall, wide, high-resolution, or text-dense images get downscaled, making text unreadable and causing AI vision to hallucinate content. This Skill tiles such images into readable overlapping crops so every region is inspected at full resolution. ## Core Features & Use Cases - Dimension Inspection: Reads width, height, and aspect ratio from the original file before deciding whether tiling is needed. - Deterministic Tiling: Splits images into overlapping tiles in vertical, horizontal, or grid mode, producing a manifest.json with exact reading order. - Overlap Reconciliation: Guides merging of tile content by deduplicating repeated boundary text and flagging unresolved OCR conflicts instead of guessing. - Use Case: Given a long scrolling screenshot of a chat conversation or a panoramic diagram, the Skill slices it into ordered tiles, reads each one, and merges the result without losing or duplicating boundary content. ## Quick Start Use the read-special-images skill to extract all text from this long screenshot without guessing any unreadable parts.

Frequently Asked Questions about read-special-images

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I read text from a very long screenshot with AI vision?▼

Slice the screenshot into overlapping vertical tiles with the slice_image.py script, then read each tile in manifest order top-to-bottom. Reconcile the 12% overlap between adjacent tiles to remove duplicated lines without dropping boundary text.

How to OCR a high-resolution image without losing detail?▼

Tile the image into smaller crops at native resolution instead of relying on the downscaled whole-image preview. Images over 4096 pixels per side or 16 million total pixels are automatically split into a grid with a reading-order manifest.

When should an image be split into tiles for vision models?▼

Tile when the aspect ratio is at least 2:1, either side exceeds 4096 pixels, total pixels exceed 16 million, or text remains unreadable in the preview. Ordinary readable images should be inspected once without tiling.

Does slice_image.py require any Python dependencies?▼

Yes, it requires Pillow for image loading, EXIF orientation correction, and cropping. If Pillow is unavailable, the script exits with a message directing you to use the existing image capability rather than installing packages without authorization.

What are the limitations of image tiling for OCR?▼

Tiling cannot recover content that is unreadable even at native resolution; those regions must be flagged as unreadable rather than guessed. Boundary words cut across tiles are only repaired when the overlap shows the complete version.

Related Skills

Explore More Software Engineering Skills→
Unson-LLCUnson-LLC
9

ttyd-upload-locator

Scan uploads and workspace directories to locate ttyd-uploaded files for attachment.

Official
Basic
aviz85aviz85
45

live-transcribe-read

Read the latest committed text from the real-time transcription log.

Community
Intermediate
waiyanphyo999waiyanphyo999
1

read-special-images

Tiles oversized or text-dense images into overlapping crops for accurate OCR and reading.

Community
Intermediate
unoplatformunoplatform
9

uno-toolkit-zoomcontentcontrol

Create a zoomable and pannable container for images, maps, diagrams, and documents.

Official
Intermediate