Flash & Photosensitivity Checker

Knowledge Base

What This Checks, and What It Cannot Tell You

Flashing imagery can trigger seizures in people with photosensitive epilepsy. Unlike most accessibility issues, the consequence is medical rather than inconvenient, which is why it is worth screening for deliberately rather than trusting your eye.

This tool implements the general flash threshold in the same shape the accessibility guidelines describe it. It decodes your video frame by frame in the browser, converts each frame to relative luminance, and looks for opposing transitions (a rise then a fall, or a fall then a rise) where the change is at least 10% of full luminance and the darker of the two states is below 80% luminance. A matched pair of opposing transitions is one flash. It then counts flashes in a sliding one-second window and reports the peak.

Three flashes in any one second is the line. More than that is the condition the guidelines treat as a failure, and the same figure appears in broadcast regulation. Treat it as a ceiling to stay well below, not a target to approach.

Reading the numbers

Peak flashes per second is the headline. Above three in any one-second window, the content fails the general flash threshold and needs changing.

Largest changing area is the greatest proportion of the frame that changed brightness significantly during a single transition. The guidelines only count flashing that occupies a meaningful portion of the visual field, roughly a quarter, so a small flickering element in a corner is much lower risk than a full-frame strobe. A high flash rate combined with a small area is worth a look but is not automatically a failure.

Saturated-red flashes are flagged separately because transitions to and from saturated red carry elevated risk beyond their brightness, and both web and broadcast guidance single them out.

Frames sampled per second tells you how thoroughly the analysis ran. If it is well below the video's real frame rate, the browser dropped frames during playback and fast flashes may have been missed, re-run at a lower analysis speed for a more reliable result.

If it fails, what to change

Most flashing effects survive being fixed, because three variables independently affect risk and you only need to move one.

Slow it down below three flashes per second. This is the most direct fix and usually preserves the intent.

Reduce the contrast of the transition. What matters is the size of the luminance change, so alternating between two mid-tones is far safer than alternating between black and white at the same rate, and often looks better.

Shrink the affected area. Confining the effect to part of the frame rather than the whole screen reduces risk substantially.

Removing the effect entirely is rarely necessary once you know which variable to adjust. If flashing is genuinely essential to the content, put a warning at the very start, before the flashing occurs, and visible rather than only in a description, since autoplay gives the viewer no chance to read anything first.

Privacy

The video is decoded locally by your own browser and analysed in the page. Nothing is uploaded, nothing is stored, and the tool works with the network disconnected. That matters here specifically: unreleased content is exactly the kind of material people cannot send to an online analyser.

FAQ

Q: What counts as one flash?
A: A pair of opposing changes in relative luminance (a rise followed by a fall, or the reverse) where each change is at least 10% of full luminance and the darker state is below 80% luminance. A single change in one direction is not a flash.

Q: Why three per second?
A: It is the threshold used by both web accessibility guidance and broadcast regulation. Risk rises sharply in a band of flash frequencies, and three per second is the conventional line below which content is treated as acceptable.

Q: My video passes. Is it definitely safe?
A: It is a good sign, not a guarantee. This is a screening aid using a simplified area and colour model, and it can only measure the frames it managed to sample. Anything heading for broadcast should be tested with a validated analyser.

Q: Why is the frames-sampled number lower than my frame rate?
A: Faster analysis speeds ask the browser to present frames quicker than it can decode them, so it drops some. Re-run at 1× or 2× for a fuller sample, slower, but it sees more frames.

Q: Does it check for patterns as well as flashes?
A: No. Regular high-contrast striped patterns can also trigger symptoms and are part of the broadcast standards, but this tool measures luminance transitions over time only.