paste-as-code

Copy a screenshot or image into clipboard, run one command, and paste back clean Markdown.

⚡ The Shortcut

pac -p -c
pac -cb -c

That is the whole magic trick:

1. Copy a screenshot or image.
2. Run pac -p -c or pac -cb -c.
3. Paste the result into ChatGPT, Slack, Discord, GitHub issues or anywhere.

🚀 Install

npm install -g paste-as-code

Or try it without installing:

npx paste-as-code --help

Requires Node.js 20 or newer.

💡 Why It Exists

Screenshots are where useful text goes to become annoying.

paste-as-code removes the tiny, repetitive steps between “I can see this” and “I can use this”:

• no manual typing
• no file-path friction
• no OCR configuration
• no copy-paste cleanup
• no Markdown wrapping
• no need to paste screenshots or images into LLMs or agents; save tokens by sending the copied text from those images instead of the image itself

📋 Clipboard First

Copy any image to your clipboard and turn it into Markdown text.

pac -p
pac -cb
pac -cb -c

That means:

• screenshot or any other image goes in on the clipboard
• OCR runs immediately
• clean Markdown comes back out
• no file paths, no dragging, no waiting around

✨ Quick Start

🖥️ Clipboard Screenshot

On macOS, copy a screenshot to your clipboard, then run:

pac -cb -c

🖼️ Image File

If you already have a file:

pac screenshot.png

The output is Markdown by default:

```ts
const value: number = 1;
```

🧭 Command Reference

Command	What it does
pac screenshot.png	OCR an image file and print Markdown.
pac -cb	Read a screenshot from the clipboard.
pac -cb -c	Read clipboard image, OCR it, and copy Markdown back.
pac screenshot.png -r	Print raw cleaned OCR text.
pac screenshot.png -l ts	Force the Markdown fence language.
pac screenshot.png -k	Keep edge-clipped lines that would otherwise be dropped.
pac screenshot.png --no-preprocess	Skip image preprocessing if it hurts a specific screenshot.

Verbose forms are still supported too:

paste-as-code --clipboard --copy
paste-as-code screenshot.png --raw
paste-as-code screenshot.png --lang ts

✅ What It Handles Well

• code screenshots
• error messages
• stack traces
• terminal output
• Discord and Slack snippets
• macOS clipboard screenshots

🎯 Accuracy Notes

The OCR pipeline is tuned for code-like screenshots:

• small images are upscaled
• dark-theme screenshots are inverted when needed
• text is normalized and sharpened before OCR
• clipped top/bottom lines can be dropped automatically
• code-friendly Tesseract settings preserve spacing and avoid aggressive dictionary correction

If a specific screenshot gets worse after preprocessing, use:

pac screenshot.png --no-preprocess

If you want to keep clipped lines anyway:

pac screenshot.png --keep-clipped

🧩 Platform Notes

• File input works anywhere the package and native dependencies install successfully.
• Clipboard image input is currently macOS-only.
• Stay tuned for Linux and Windows clipboard support.

🛠️ Development

npm install
npm run typecheck
npm test
npm run build

Manual OCR smoke test:

npm run test:ocr -- screenshot.png

📦 Release

This repo is published to npm as paste-as-code.

If it saves you time, starring the repo helps other developers find it too.