Importing a Long Page Into Figma: One Full-Page Capture, Layer Caps, and Lazy Images
How to do a long page Figma import in one full-page capture: the 500 layer free cap, lazy-loaded images, and the Figma limits that matter.
For a long page Figma import, capture the whole page once, at one width, and import it in a single pass. UnHTML has no documented mode for importing a page in sections, so this article does not invent one. What does matter on a tall page is the layer cap on the free tier, images that never loaded because nobody scrolled, and a few Figma limits.
This tutorial is for developers with a long marketing page, docs page or storefront they want as editable layers. You need the UnHTML plugin in Figma and a captured copy of the page. The steps run from capture to a clean read of the import report.
Why not capture a long page in stitched sections
The tempting workaround is the screenshot habit. Take the page in four or five slices, drop them into Figma, line them up. It fails for two reasons. The first is resolution. The Figma plugin API states that "images have a maximum size of 4096 pixels (4K) in width and height", so a single tall screenshot of a long page cannot stay at full resolution, and slices leave you with seams to hide.
The second reason is that slices are pictures. You get no text layers, no frames and no auto layout, which is the whole point of importing a page instead of tracing it. We compared the two approaches in what a screenshot trace gives you.
A single DOM capture avoids both problems. Every section shares one coordinate system, one set of styles and one import report, so the footer lines up with the hero because they came from the same page.
Capturing the full page in one pass
Prerequisite: the UnHTML plugin installed in Figma and a captured file of the page. UnHTML accepts three kinds of input, and the right one depends on how the page is built.
- Pick the input format. The plugin takes a self-contained
.htmlfile, a Safari.webarchive, or ascene.jsonfrom the open-source CLI or bookmarklet. Self-contained files and web archives carry their own assets, so they need no outside requests. Live sites, remote stylesheets and pages built by JavaScript are best captured toscene.jsonfirst. - Use the CLI for pages that need a browser. The CLI runs a headless browser, and the UnHTML site describes it as doing a full-page scroll. A typical call looks like this:
node capture/cli/capture.js https://example.com --out scene.json. The site does not say whether the bookmarklet also scrolls, so for a long page prefer the CLI. - Choose one width. The import step offers Desktop 1440, Mobile 390, or a custom width. Pick the width you are redesigning for. A long page captured at 1440 and again at 390 gives you two separate imports, which is the honest way to compare layouts.
- Import once. Run the import on the full file, not on pieces of it.
Nothing here makes the result a pixel-exact clone. It is a faithful, editable set of layers, and the import report tells you where it fell back.
Lazy-loaded images
Long pages are where lazy loading lives. MDN explains that the loading attribute defers loading of off-screen resources until the user scrolls near them, and it applies to images, iframes, video and audio. MDN also warns that when the load event fires, "it's entirely possible (or even likely)" that lazily loaded content is still waiting.
That matters if you save a long page by hand. If you export the file before you have scrolled to the bottom, images far down the page may never have been fetched, and they will be missing or empty in the capture. The UnHTML site does not discuss lazy images, so treat this as a reasonable precaution rather than a documented behavior:
- Scroll the whole page slowly before you export a self-contained HTML file or a web archive.
- Prefer the CLI capture for long pages, since it does a full-page scroll.
- After the import, open the report and check which images were skipped.
The 500 layer cap and when it matters
A long page has more elements than a short one, and layers add up quickly. As of 2026-11-08, the UnHTML site lists the free tier at up to 500 layers per import. Pro is $39 one-time per editor and has no layer cap per import, described as the tier for long, dense pages in one pass.
The site also mentions a node budget that guards against runaway imports, but publishes no number for it, and it states no maximum page height. So the practical rule is simple: if a long page runs past 500 layers on the free tier, the answer is the Pro tier, not slicing the page. If you are weighing the two, which UnHTML tier fits covers what each one includes.
Figma limits to keep in mind
Figma's help center says that "pages have a finite size of around -130,000 through +130,000 on each axis". That's a large canvas, and a long web page fits comfortably. The same article states no limit on the number of objects, frames or layers, so the cap you will run into on a long page is the plugin's, not the file's.
Remember the image limit from earlier too: 4096 pixels per side applies to any bitmap on the page, however the page got there.
Reading the import report on a long page
After the import, the report lists three things worth reading closely on a tall page: font substitutions, frames that reverted to absolute positioning, and images that were skipped. Long pages repeat the same patterns many times, so one missing font can show up in dozens of text layers. Fix it once and the rest follows.
Skipped images are your lazy-loading check. If a section near the bottom has empty image frames, go back, scroll the page fully and capture again. Once the layers look right, deleting the noise after an import keeps a long file easy to work in.
A checklist before you press import
Long pages punish small mistakes. One bad capture costs you a full import to discover. Run through this short list first:
- Same width every time. Decide on 1440, 390 or a custom width before you capture, and write it down. If you later re-capture a section, a different width changes the whole layout, so the two imports will not match.
- A fully scrolled page. Reach the footer before you save, so every lazy-loaded image has had a chance to load.
- The right input for the page. A static page can go through a self-contained
.htmlfile or a Safari.webarchive. A page built by JavaScript goes through the CLI toscene.json. - A known layer budget. If you expect more than 500 layers, plan on Pro before you start, so you are not forced to cut the page into pieces after the fact.
- A place to put the result. Importing into a fresh Figma page keeps a very tall frame away from the rest of your file.
If you still need to work on one part of a long page, import the whole thing and then work on the part you want. You can select and copy a section of the imported layers inside Figma, which beats stitching slices and keeps the original structure intact.
Conclusion
Capture once, at one width, and import once. Scroll the page before you save it, check the report for skipped images, and look at the 500 layer count if you are on the free tier. Slicing a long page into sections adds seams and image limits without adding anything the plugin can use.
If you want to try this on your own long page, start from the UnHTML home page.
Questions
- Does UnHTML have a maximum page height?
- The UnHTML site states no maximum page height and publishes no number for its node budget. The free tier stops at 500 layers per import, and Pro has no layer cap per import.
- How many layers does the free tier import from a long page?
- Up to 500 layers per import. Pro, at $39 one-time per editor as of 2026-11-08, has no layer cap per import.
- Why are some images missing after I import a long page?
- Browsers defer lazy-loaded images until the user scrolls near them, so a page saved without scrolling can lack the images far down. Scroll the whole page before saving, or use the CLI capture, which does a full-page scroll, then check the import report for skipped images.
- How tall can a Figma frame be?
- Figma documents no maximum frame height. It does say pages span around -130,000 through +130,000 on each axis, and the plugin API limits images to 4096 pixels in width and height.