Pulling numbers back out of a graph has always meant the same tedious ritual: mark the axes, then click every single data point, one at a time, hoping your hand stays steady along the curve. A dense plot can cost twenty minutes of clicking.
Plot Extractor is built the other way around — automation first, manual second. Drop in a plot and it does the first pass itself: it finds the plot region, places the axis anchors, reads the tick labels, picks out each curve by colour, and traces the points. What lands in front of you is not a blank canvas — it's a finished extraction you only need to check.
It is, as far as we know, the first free plot digitizer designed this way. Older tools treat automation as an optional extra bolted onto a manual workflow. Here it is the workflow.
The fastest way to see it: copy any plot image to your clipboard and paste it straight onto plotextractor.com. No file dialog, no upload step, no account. A screenshot from a paper, a figure from a datasheet, a chart from a slide — paste it and the first pass runs immediately.
The moment a plot loads, Plot Extractor runs a chain of detectors so the boring setup is already done:
By the time the screen settles, you are looking at extracted data, not a setup wizard.
Automation gets most plots most of the way. It will not get every plot perfectly — a curve crossing a gridline, two traces touching, a dashed line, a noisy scan. That is expected, and it is exactly why the manual tools exist.
The key difference: you correct the automatic result, you never start from zero. If one curve picked up a stray gridline, you fix that curve. If a few points drifted, you nudge those points. The 95% the tool got right stays put. Manual work is reduced to touch-ups.
Three precise tools handle almost every correction. They are the same tools shown in the in-app demo tour.
When a plot has several curves, point the colour picker at the one you want. Plot Extractor isolates every pixel of that colour and traces just that series — no manual outlining.
The pen mask is a brush for the trace itself. Paint over a gridline the tracer mistook for data, or paint back in a faint stretch of curve it missed. The trace updates as you brush.
For the spots that need a human eye — a sharp knee, a crossing, the very end of a curve — switch to manual point editing and place or drag individual points. Pixel-level control, only where you actually need it.
That is the whole loop: paste a plot, let the first pass run, correct the few things that need it, export to CSV or JSON. Automation carries the load; you supply judgement. The tedious part is gone.