What "spread" actually asks of a renderer
Figma now supports adjusting shadow spread on rectangles, ellipses, frame backgrounds, and component backgrounds, matching the behavior of CSS box-shadow. The feature was first requested on Spectrum 958 days before it shipped, and the gap is explained largely by how deceptively simple the request sounds: a spread value is the distance by which a shadow expands or contracts in all directions. Scaling the shadow's geometry gets you there for a rectangle, but not for shapes with holes, such as the Figma logo.
Figma's existing shadow pipeline copies the object's geometry, fills it with one color, blurs it, and draws it beneath the node. Doing "the right thing" for spread—extruding the shape outward in every direction—had no clean fit in the renderer, and the workaround of emulating spread with strokes failed as well: certain vertex angles are handled differently in strokes, and the prototype renderer contains no stroke generation code. Whatever route was chosen could not add large amounts of geometry code to two separate rendering codebases.
Letting the CSS spec set the scope
The turning point was reframing what the feature is for. Users currently ship workarounds, and separately maintained documentation for developer handoff, whenever shadow spread is involved. Since the goal is to reduce friction at handoff, CSS should define the constraints—including whether a perfect spread shadow of a hole-y shape is even necessary.
It isn't possible in CSS anyway. box-shadow renders shadows of boxes (and box-like shapes, including ellipses when corner radii are chosen correctly); asked to shadow the Figma logo, it draws a box rather than a blobby copy of the logo. Note also that spread in filter: drop-shadow() is not a thing: MDN pages suggesting the spec included it were mistaken, as the W3C spec states outright.
Designer advocates on the team were confident that rectangles and ellipses alone would cover a large share of use cases. With that reassurance and CSS compliance as the guiding principle, the team scoped the work to shapes where box-shadow applies and generated a larger or smaller version of the original geometry—not a simple stretch, which would break rounded corners, but new rectangle geometries that both rendering engines can produce.
Ellipses, corner radii, and just copying the browser
Several subproblems surfaced. A true spread shadow of an ellipse is no longer elliptical, yet the ellipse generation code only produces ellipses, and transforming an ellipse in either direction preserves ellipticity. Per the W3C spec, applying spread to a rounded rectangle involves both a general rule for transforming corner radii and a specific formula for large spread values. And stroke-only nodes have no CSS-defined behavior to match.
Some of these were resolved by testing in a CodePen and observing what browsers do. Browsers implement elliptical shadows with spread by generating a larger ellipse rather than a spread blob, and since the team had committed to following CSS, Figma does the same. For corner radii, however, applying the specific W3C cubic rule for large spread values and comparing against a quick CodePen revealed that browsers, as of publication, don't implement that rule at all. Both browsers and Figma instead simply add or subtract the spread value and the original corner radius.
Strokes: four readings, one choice
CodePen offered no guidance for stroke-only nodes, because Figma's shadow model already diverges from CSS here. CSS lets fully transparent fills factor into and mask their own shadows (though not other shadows); Figma leans closer to physically-based rendering, showing shadows through transparent and translucent fills and excluding invisible geometry from shadows.
With a fill, rendering a shadow is straightforward. With a stroke, spread could plausibly be interpreted four ways:
- Outset the stroke by
spread, leaving the stroke constant. - Add
spreadto the outside of the stroke width. - Add
spreadto the stroke width, centered so it is distributed on either side. - Add
spreadto each edge of the stroke width, ultimately adding2 * spreadto the stroke width.
Option D won. The deciding consideration was that toggling the object's fill visibility should leave the shadow's outer footprint unchanged, which ruled out C; among the rest, D best matches the notion of a shadow extruded along every point by spread.
Lesson taken
The path from a Maker Week afternoon project to a shipped feature involved algorithmic exploration, spec reading, and product tradeoffs. The takeaway offered: when interpreting a feature request, think about the motivation behind it, and study the tradeoffs made by those who came before. A playground file accompanies the release.



