Plotlift

Extract numbers from images of plots. Plotlift is a browser-based plot digitizer in the spirit of g3data: open a picture of a graph, tell it where the axes are, click on the data, and export the values as text.

Use it here: https://aldorvasq.github.io/plotlift/

It runs entirely in your browser. Nothing is installed, and your images and data never leave your computer.

Features

• Opens PNG, JPEG, GIF, WebP, BMP and SVG images, pages from PDF files, and pasted screenshots. Several plots can be open at once in tabs.
• Calibrates axes from four reference points (X1, X2 on the X axis, Y1, Y2 on the Y axis). Each axis can be linear or logarithmic. Rotated or skewed scans are handled.
• Magnifier with a live X/Y readout for precise clicking. Optional snapping to marker centres, crosshair guides, and a calibrated grid overlay to check the calibration against the plot's own gridlines.
• Auto-detect finds data points by colour: click the data to pick its colour, outline the search area, and Plotlift places a point at the centre of each marker (splitting touching markers) or traces a line at a spacing you choose. The search area can be drawn as a box, a polygon or a freehand shape, and shapes combine: hold Shift to add to the area or Option/Alt to cut a piece out of it, for example to leave out a legend that sits inside the plot. You review the proposed points before adding them.
• Several data series per plot, each with its own colour. Points can be dragged, nudged with the arrow keys, deleted, undone and redone.
• Exports points in click order or sorted by X or Y, as space-, tab- or comma-separated text, with optional error columns (dx, dy). Copy to the clipboard or save as .txt or .csv.
• Exports one series or all of them at once. With all series you choose the layout: separate blocks (gnuplot style), one table with a series column, or side by side with an x and y column pair per series (handy for Excel or Origin). Save each series as its own file packs one file per series into a .zip.
• Saves and reopens whole projects (images, axes and points) as a single .json file.

How to use it

1. Upload a plot. Plotlift opens on an upload screen: drop an image or PDF onto it, click Choose a file…, or paste a screenshot. To see a finished example first, click Open the sample plot.
2. Calibrate. Follow the prompts: click a known point on the X axis and type its value, then a second one, then two on the Y axis. Choose Linear or Log for each axis.
3. Mark points. Click on the data. Use the magnifier on the right for precision, and drag points to adjust them. Or switch to Auto-detect (press D): click a marker or the line to pick its colour, then outline the data so the axes, labels and legend are left out. Choose Box (drag), Polygon (click each corner, then click the first corner or double-click to close) or Freehand (drag around the data). Hold Shift to add another shape to the area, or Option/Alt to cut one out. Check the crosses it proposes, then click Add points. Detection works best when the data has its own colour; data drawn in black like the axes needs a tight box and some manual clean-up.
4. Export. Pick the order, separator and columns, then copy or save.

Press ? inside the app for the full list of mouse and keyboard shortcuts.

Running it locally

Download Plotlift.html and open it in any modern browser (Chrome, Edge, Firefox or Safari from 2023 or later) on Windows, macOS or Linux. It works offline, except that opening a PDF needs an internet connection the first time, to load pdf.js.

Making changes

The app is a single file of HTML, CSS and plain JavaScript with no frameworks or build tools.

• Edit src/plotlift.html.
• Run ./build.sh. It wraps the source into the standalone pages index.html (served by GitHub Pages) and Plotlift.html (for downloading).
• Open Plotlift.html in a browser to check your change, then commit and push. The website updates within a minute or two.

The test/ folder holds plots generated with matplotlib whose true values are recorded in test/fixtures.json, for checking that extracted values are accurate: a log-log scatter, a rotated scan, a continuous curve for line tracing, and a scatter with touching markers.

How the numbers are computed

The X value of a point is read along the line from X1 to X2, and the Y value along the line from Y1 to Y2, so the axes need not be level or perpendicular. On a log axis the interpolation runs in log₁₀ space. The optional dx and dy columns propagate a click uncertainty (half an image pixel by default) through the calibration.

License

MIT © 2026 Aldo Vasquez-Briceno