Check whether an animation is safe to ship. The main test is for flashing: rapid changes in brightness can trigger seizures in people with photosensitive epilepsy, and WCAG sets the limit at three flashes in any one second. Measuring that properly means rendering the animation rather than reading its JSON, so this page draws every frame to an offscreen canvas and measures relative luminance the way WCAG defines it. It also reports motion intensity — a separate problem, since large sustained movement causes nausea and dizziness for people with vestibular disorders without any flashing at all — along with duration, file weight, embedded rasters, frame rate and expressions. Every result explains its reasoning, and there is a reduced-motion snippet to copy. Everything runs in your browser and the file is never uploaded.
| Check | Threshold | Who it affects |
|---|---|---|
| Flash rate | More than 3 flashes in any 1 second | People with photosensitive epilepsy |
| Motion intensity | Large sustained frame-to-frame change | People with vestibular disorders |
| Duration | Over 5 seconds, or looping | Anyone who needs to stop it (WCAG 2.2.2) |
| File size | Warn over 150 KB, fail over 500 KB | Slow or metered connections |
| Embedded images | Any raster asset | High-contrast and dark themes |
| Frame rate | Over 60 fps | Low-end devices |
| Expressions | Any present | Playback cost and player support |
It renders the animation frame by frame and measures the mean relative luminance of each frame, then looks for pairs of opposing changes of at least ten percent where the darker state is below 0.8. More than three of those in any one second is the WCAG 2.3.1 failure point. It has to render rather than read the file, because one shape layer with an animated fill can strobe harder than fifty layers of gentle motion.
Because it does not trigger photosensitive responses. The threshold requires the darker of the two states to be below 0.8 relative luminance — bright-on-bright flicker is visually annoying but not the hazard this rule addresses.
No. It is a heuristic against a published threshold, measured on a rendered sample. It will catch obvious strobing and give you a reasoned starting point, but a formal accessibility audit involves judgement this cannot make — particularly around the size of the flashing area, which WCAG also accounts for and this does not.
They harm different people. Flashing is a seizure risk driven by luminance reversal. Motion intensity is about large sustained movement, which causes nausea and dizziness for people with vestibular disorders and involves no flashing at all. An animation can pass one and fail the other, so they are reported separately.
WCAG 2.2.2 requires that anything moving automatically for more than five seconds can be paused, stopped or hidden. A looping animation is moving indefinitely, so the rule applies to it however short one cycle is.
Raster assets dominate the file size, do not scale cleanly, and cannot be recoloured — so an animation that embeds them cannot adapt to a high-contrast or dark theme the way a pure vector one can.
No. It is rendered and measured in your browser and never sent anywhere.
The flash test cannot be done from the JSON, because what matters is what the animation looks like, not what it is made of — a single shape layer with an animated fill can strobe harder than fifty layers of gentle motion. So the animation is rendered frame by frame to an offscreen canvas and each frame is reduced to its mean relative luminance. Relative luminance is WCAG's definition, not a channel average: sRGB values are gamma-encoded and must be linearised before weighting, and green counts for roughly seven times what blue does, because that is how human vision responds. Averaging the raw bytes instead would misjudge exactly the saturated colour flashes the test exists to catch. Fully transparent pixels are excluded rather than counted as black, or every animation on a transparent background would appear to flash against its own emptiness. Frames are sampled at the file's own frame rate, since the threshold is defined per second of real time — sampling coarser misses flashes, sampling finer invents them. The sequence is then measured against WCAG 2.3.1: a flash is a pair of opposing luminance changes of at least ten percent where the darker of the two states is below 0.8, because two bright shades alternating do not trigger photosensitive responses however fast they alternate, and more than three flashes within any one-second window is the failure condition. That window is a sliding one rather than an average over the animation, which matters more than it sounds: one violent second inside ten calm ones is precisely the case that causes harm, and an average would dilute it to nothing. Motion intensity is reported separately from flashing, because vestibular triggers are about large sustained movement rather than luminance reversal and an animation can be dangerous on one axis while fine on the other. The remaining checks read the file directly — duration against WCAG 2.2.2's five-second pause requirement, weight, embedded rasters that cannot be recoloured for high-contrast themes, frame rates beyond what displays show, and expressions that cost CPU on every frame and are unsupported by several players. Every check states its reasoning, since a report that says "fail" without saying why leaves the reader no better off than before they ran it.