What Snapture actually produces
Marketing pages show mockups. This page shows measurements. We ran Snapture's converter against six public pages and counted what came out the other side.
| Page | Text layers | Vectors | Images | Auto Layout | Total nodes | Depth |
|---|---|---|---|---|---|---|
| Linear linear.app | 554 | 215 | 52 | 261 | 2082 | 20 |
| Stripe stripe.com | 489 | 153 | 88 | 99 | 1592 | 23 |
| Tailwind CSS tailwindcss.com | 398 | 67 | 110 | 89 | 1540 | 18 |
| GitHub github.com | 153 | 72 | 90 | 42 | 785 | 15 |
| Notion notion.com | 92 | 22 | 58 | 31 | 389 | 11 |
| MDN Web Docs developer.mozilla.org | 116 | 7 | 19 | 5 | 253 | 8 |
Measured 30 August 2026 at a 1440×900 viewport, above the fold. Your numbers will differ — these pages change.
Full-page run, 1 September 2026
Same converter, harder conditions: the entire page this time, not just above the fold, captured at a 1280px viewport by the extension's own batch mode and counted from the saved .figma.json. Every image arrived inlined — zero captures contained a remote URL, which is what "nothing depends on the network afterwards" looks like in practice.
| Page | Total nodes | Text layers | Auto Layout | Design tokens read |
|---|---|---|---|---|
| Stripe stripe.com | 2073 | 629 | 164 | 0 |
| Tailwind CSS tailwindcss.com | 1533 | 391 | 99 | 200 |
| GitHub github.com | 773 | 150 | 48 | 3 |
| MDN: the flex reference developer.mozilla.org | 3148 | 1463 | 111 | 39 |
"Design tokens read" is how many CSS custom-property colour names the capture recovered — on Tailwind's own site, 200 of its oklch palette variables came through under their original names and become Figma variables on import. These tables are re-measured when the converter changes; the date in the heading is the measurement date.
What the columns mean
Text layers are real Figma text nodes, with the font, size, weight, letter-spacing and colour the browser computed. Not images of text. Linear's homepage alone produced 554 of them, each one something you can retype in Figma.
Vectors are SVG icons and illustrations that came through as editable vector paths rather than flattened pictures.
Auto Layout is the count of frames where the converter decided Figma's layout engine could reproduce the browser's positions within half a pixel, so it attached a layout instead of absolute coordinates. Everywhere else it kept exact coordinates on purpose.
Read the Auto Layout column carefully
Linear's page produced 261 Auto Layout frames; MDN's produced 5. That is not Snapture working better on one site than the other. It is the gate doing its job: MDN's layout is largely document flow and floats, which Figma's engine would not reproduce identically, so the converter declined and used exact absolute positions instead.
A converter that attached Auto Layout to all of them would score better on a feature checklist and hand you a file that has quietly moved. We would rather report five.
How to reproduce this
Install the extension, open any of those pages, and capture. The counts are visible in Figma's layers panel after you paste. There is nothing to take on trust here — these are public pages and the converter runs on your machine.
If you want to check the method rather than the result: the converter is dom-to-figma.js in the extension package, which you can read after installing, and the same function was called here with no arguments beyond the page body.
What this does not show
Node counts are not fidelity. A page can produce thousands of correct nodes and still get one gradient wrong. These numbers say the converter reaches deep into real, complicated pages and comes back with structure rather than a flat picture; they do not promise every pixel is perfect, and no honest measurement could.
The most useful test is the one on your own pages. The free plan is five captures with every feature on, which exists precisely so you can run that test before paying anything.
See what's included