Screenshot OCR Text Recognition

Problem It Solves

When converting meeting screenshots, contract images, report screenshots, or chat images into editable text, the hard part is not simply reading characters. It is preserving structure: Chinese and English text, numbers, code, tables, paragraph breaks, and line endings. Copy-paste does not work on images, and manual transcription can easily split tables, merge paragraphs, or disrupt mixed-language content. This skill turns such requests into a focused image OCR workflow.

How It Works

The skill is designed around image file paths, supporting both quick single-image extraction and batch processing:

• Image reading: supports common formats such as .png, .jpg, .jpeg, .bmp, and .webp.
• Text extraction: recognizes mixed Chinese-English content and preserves original paragraph structure and line breaks where possible.
• Batch processing: processes multiple images in order, separates results with dividers, and labels the source file names.
• Table recovery: detects tables and chooses the output format; simple tables become Markdown tables, while complex or data-heavy tables are exported as CSV.

The key flow is: user provides image paths -> the skill reads images with Read -> it extracts text or tables based on content -> it formats output for single or batch results.

Boundaries

It is suited for extracting visible text from screenshots, scans, or table images, not for guessing invisible or missing content. Blurry, low-resolution, occluded, or heavily handwritten images may produce unusable results, so a clearer source image is required. For highly complex layouts, multi-page tables, or dense data, review the output manually before using it in downstream document pipelines.