Rebuilding a Legacy Page in Figma, Starting From an Import

Import the old page, read the report as a punch list, then rebuild the layout in auto layout frame by frame.

TUTORIALDESIGNERPublished

To redesign a legacy page in Figma, import the page as it renders today, read the import report before you touch the canvas, and then rebuild the layout frame by frame in auto layout. Do not try to polish the import into the final design. It is a measured reference, not an exact clone of the page.

This tutorial assumes you have a saved copy of the old page and the UnHTML plugin installed in Figma. It covers why a legacy page is awkward to import, what to read in the report, what to throw away, and how to swap each old layout trick for a frame you can edit.

What makes a page legacy, and why that matters for an import

A legacy page, for this purpose, is one whose layout was built with tools that predate flexbox and grid. The usual suspects are floats, clearfix hacks and nested wrapper elements that exist only to push things into place.

The MDN guide to floats explains the history. The float property was introduced so text could wrap around an image, "the kind of thing you might get in a newspaper layout." Developers later floated entire columns to build page layouts, and MDN now says that "using floats in this way should be regarded as a legacy technique."

The MDN page on legacy layout methods names the structural weakness. A floated grid deals with columns and spanning across columns, but not rows. It is very difficult to control the height of elements with these older methods without setting an explicit height.

That matters for an import. A float layout often produces geometry that has no clean auto layout equivalent: columns of unrelated height, offsets made with negative margins, and wrappers three levels deep. Expect more frames to fall back to absolute position than on a modern flexbox page.

Capture the page as it renders today

Capture the page the way it looks in a browser now, not the way the template files describe it. UnHTML accepts a self-contained .html file, a Safari .webarchive, or a scene.json from the open-source CLI or bookmarklet. The CLI is the right choice for pages built with JavaScript or pages that pull remote assets. The same capture steps apply when you are importing a WordPress theme page, which is a common kind of legacy page.

With default settings the import makes zero network requests. Scripts, inline event handlers and javascript: URLs are stripped from the markup before it renders. Fetching remote assets is an opt-in toggle that is off by default.

Check the size before you start. As of 2026, the free tier imports up to 500 layers per capture. A long legacy page with a heavy footer and a mega menu can pass that. Pro removes the cap for $39 one time per editor.

Step 1: read the import report first

Every import produces a report. It covers three things, and for a legacy page each one tells you something different.

  1. Font substitutions. Fonts that are not installed on your machine are replaced, with Inter as the default, and every swap is listed. Old sites often name a font stack that starts with a font nobody has installed. Write down each substitution and decide whether the redesign keeps that typeface at all.
  2. Frames reverted to absolute. A frame lands here when auto layout cannot reproduce the captured geometry. On a legacy page this list is your map of the old layout hacks. Each entry is a place where the original author used a float, a negative margin or a stack of wrappers to get a result that flexbox would express in one rule.
  3. Skipped images. These are images the import did not bring in. Compare them against the live page and note which ones the redesign actually needs.

Treat the report as a punch list, not a log. The next steps work through it.

Step 2: decide what to keep

Before you rebuild anything, delete what the redesign will not use. The import captures what the page rendered, including things a design file should never carry.

  • Remove cookie banners, chat widgets and floating share buttons.
  • Remove spacer elements and empty wrappers that only existed to hold a float in place.
  • Compare the canvas against the live page at the same viewport width, and check line breaks in headings and buttons.
  • Resolve the font substitutions from the report: pick the real typeface, or choose a replacement on purpose.

Deleting first keeps the layer list short. If you want a longer routine, the article on how to clean up the import covers what to delete and what to flatten.

Step 3: rebuild the layout in auto layout

Now replace each absolute frame from the report with a frame that uses auto layout. Work from the outside in, one section at a time.

  1. Pick the container. Take one section, such as the header or the three-column feature row, and select its outermost frame.
  2. Set the direction. Decide whether the section is a row or a column. A row of floated columns becomes a horizontal auto layout frame. This maps to flex-direction in CSS terms.
  3. Read the spacing off the import. The imported layers carry the measured positions. Use the distances between neighbors to set gap and padding, and round them to your spacing scale.
  4. Move children into the frame. Drop each child in and set whether it hugs its content, fills the space or has a fixed size.
  5. Keep absolute position only where layers genuinely overlap. A badge on a card corner or a decorative shape is a legitimate use. A column that is merely offset is not.

The reason this works is that flexbox was designed for the job floats were stretched to do. The W3C flexbox specification says flex layout lets an author trivially achieve effects that previously required complex or fragile methods, such as hacks using the float and clear properties. Flexbox is one-dimensional, so a section that needs rows and columns together is usually two nested frames, or a grid-like arrangement you build by hand in Figma.

You don't have to rebuild every pixel. Rebuild the structure, then adjust spacing, type and color in the design system you actually want to ship.

Keep real data tables as tables

Old pages sometimes used HTML tables for page layout. In the redesign those become ordinary frames. A table that holds real tabular data, such as a price list or a spec sheet, is different: keep it as a table in your spec and in the build.

The MDN reference for display warns that in some browsers, changing the display value of a table element to block, grid or flex alters how the table appears in the accessibility tree, so screen readers no longer announce it properly. Note this in your handoff so developers do not turn a data table into a flex container.

Conclusion

The sequence is short: import the page as it renders, read the report, delete what the redesign will not use, then rebuild each frame the report flagged in auto layout. The report is your punch list, and the absolute frames in it point to the old layout hacks. UnHTML gets you the measured starting point, but the redesign is still your work, and that is the right split for a page that was built before flexbox existed.

Frequently Asked Questions

Can I redesign a legacy page in Figma if I have no design file?

Yes. Import the rendered page, read the report, and use the layers as a measured reference to rebuild from. The import is a starting point, not an exact clone of the page.

Why do some frames in a legacy page import come in with absolute positioning?

Frames whose layout cannot be reproduced with auto layout fall back to absolute positioning. Float layouts and deeply nested wrappers are common causes. Every fallback is listed in the import report.

Do I need the paid tier to import an old page?

Not necessarily. As of 2026 the full importer is free up to 500 layers per import. Pro costs $39 one time per editor, removes the cap, and adds components and shared text styles.

Does importing a legacy page send it anywhere?

With default settings the import makes zero network requests, and scripts are stripped from the markup. Fetching remote assets is an opt-in toggle that is off by default.

Questions

Can I redesign a legacy page in Figma if I have no design file?
Yes. Import the rendered page, read the report, and use the layers as a measured reference to rebuild from. The import is a starting point, not an exact clone of the page.
Why do some frames in a legacy page import come in with absolute positioning?
Frames whose layout cannot be reproduced with auto layout fall back to absolute positioning. Float layouts and deeply nested wrappers are common causes. Every fallback is listed in the import report.
Do I need the paid tier to import an old page?
Not necessarily. As of 2026 the full importer is free up to 500 layers per import. Pro costs $39 one time per editor, removes the cap, and adds components and shared text styles.
Does importing a legacy page send it anywhere?
With default settings the import makes zero network requests, and scripts are stripped from the markup. Fetching remote assets is an opt-in toggle that is off by default.