Screenshot Magician — Capture & Annotate

Screenshot tabs and combine them into one labelled comparison image. Annotate, export, or copy — offline, no account, no tracking.

Screenshot Magician captures tabs and gathers them into one labelled image, so a change is something to look at rather than describe. CAPTURE Click the toolbar icon on a tab and it takes a fresh shot. Tick "Continue adding" and each tab you switch to joins the set — up to ten captures. Right-click a page or an image to add it from the menu, or gather a set in Chrome's side panel without leaving the page. ARRANGE - Combined: one labelled image — side by side, stacked, or in a grid up to four wide. - Separate: one WebP per capture, at its full resolution and nothing rescaled. - Per row, gap, border, shadow, and rounded corners. - A label per capture: text, band colour, text colour, and four styles — bar, text, tab, chip. - Resize to any percentage. Labels and annotations are drawn at that size rather than shrunk into it, so they stay sharp. - Swap the two captures when they were taken in the wrong order. ANNOTATE Arrow, line, box, text, numbered marker, and mask — which pixelates what is under it, for hiding names and account numbers — plus crop. 18 colours, five line weights, three fonts, bold, a white outline, and a drop shadow. Undo, clear all, select a shape to move or resize it, copy and paste it, and a live readout of its size while you draw. EXPORT - Download into Downloads/screenshot-magician/, or into a folder you pick and it remembers. - Copy to Clipboard, for a pull request, an issue, or a chat. HAND OFF TO A MODEL Copy for AI saves the render and puts the image and a JSON description on the clipboard together, ready to paste into Claude, ChatGPT, Grok, or an agent that can open files itself. Each capture carries its absolute file path, page title, URL and capture time, and every shape its type, position, size, colour and text — in pixels of the exported image, so a model can point at exactly what it means. The description is one minified line, to cost as few tokens as the content allows. No model is built in and nothing is sent anywhere: the clipboard is yours, and what happens to it next is whatever you paste it into. HISTORY Every capture is kept locally, newest first, with a filter, four thumbnail sizes, and grouping by week or by domain. PRIVATE BY CONSTRUCTION - No network requests at all. Nothing is loaded from a remote host — no CDN, no web font, no external script. - Nothing is uploaded. No telemetry, no analytics, no account, no sync. - Captures, settings and annotations live in this browser profile and never leave it. PERMISSIONS activeTab screenshots the tab you invoked the extension on. clipboardWrite puts the rendered image on the clipboard. downloads writes exported files. contextMenus adds the right-click menu. sidePanel shows the comparison view in Chrome's side panel. None of them grants access to a page you have not invoked the extension on, and "Allow any tab" is an optional permission you can refuse. WORTH KNOWING Captures the visible viewport, not the full scrollable page. Chrome allows two captures a second, so rapid clicks are throttled. Screenshots come back at your display's pixel ratio.