---
name: Snapping
slug: snapping
category: interaction
status: published
created: 2026-08-21T00:00:00.000Z
modified: 2026-08-21T00:00:00.000Z
definition: Pulling a dragged object into exact alignment once it comes within a
  small distance of a grid, guide, or another object's edge.
aliases:
  - name: snap to grid
    source: community
  - name: magnetic snapping
    source: community
  - name: snap to object
    source: community
  - name: snap tolerance
    source: community
tags:
  - canvas
  - dragging
relations:
  contrastWith:
    - smart-guides
    - drop-indicator
    - detent
  variantOf: []
  partOf: []
  seeAlso:
    - drag-and-drop
    - nudge
implementations: []
sources:
  - title: "Wikipedia: Snap (computer graphics)"
    url: https://en.wikipedia.org/wiki/Snap_(computer_graphics)
demo: inline
exhibit: false
useWhen: a drag jumps into alignment near a guide
---

Nobody can drop a box on an exact coordinate with a mouse, and no design would be
better for it if they could. Snapping takes the intention and discards the tremor: while
the object is inside a tolerance of some reference, its position is quietly replaced by
the reference's own. What it snaps to is the interesting part, because the references
form a hierarchy. A fixed grid is the coarsest. Guides a person placed themselves are
next. Then the edges, centres, and baselines of the neighbouring objects, which is what
makes a diagram tool feel like it understands the drawing rather than just the canvas.

The tolerance is measured in screen pixels, not in document units, which is the detail
people get wrong first. Snap at a fixed model distance and the behaviour changes
character as the reader zooms: irresistible when zoomed out, unreachable when zoomed in.
Keep the catch radius constant on screen, somewhere around 5 to 10 pixels, and it feels
the same at every scale. The other classic mistake is snapping the pointer instead of
the object. The reader is dragging a shape, so what has to land on the line is the
shape's edge or centre, and every candidate the shape offers should be tested against
every nearby reference so that a corner, a midpoint, or a baseline can each be what
catches.

Show your work while the drag is live. Snapping that only happens on release is a
lottery, so tools draw the alignment as it is found: the guide the object is about to
lock onto lights up, an outline shows where the drop will land, and something (a
distance badge, a matching gap marker) says why that position was chosen. A snap that
happens silently and invisibly reads as the object refusing to go where it was put.

Which brings up the escape hatch. Snapping is a guess about intent, and it is sometimes
wrong, so every serious tool lets a reader overrule it by holding a modifier during the
drag or turning the grid off outright. Without that, the last few pixels of a layout
become impossible, and the feature that was supposed to make alignment effortless is
the thing standing in the way.