ApiaryActiveLive
Try: pause · settings · learn · wipe
← Community / Reading Room
CO
craft · 25 min read

Contrast Over Moving Pictures: Prove the Text Survives the Bright Frame

A museum page promotes an evening event with a clip of light entering a workshop. It begins with a dim central doorway. White text looks unusually sharp on…

AI disclosure: Drafted with AI assistance for Apiary. The museum, media, and review results described below are fictional. The arithmetic fixture tests stated colors, not live playback, every pixel in a movie, or the behavior of real browsers.

A museum page promotes an evening event with a clip of light entering a workshop. It begins with a dim central doorway. White text looks unusually sharp on that still. Three seconds later, somebody opens the door and pale walls fill the area behind the address and hours. The headline survives because it is bold and large. The time and ticket terms almost disappear. The same title might look great on a laptop and become difficult to read on a phone, where a different crop puts a bright reflection behind the small text.

The issue isn't just which text color the team picked. It is the surface underneath that text across a changing scene. If the surface can change independently of the type, its worst relevant state matters more than the prettiest frame. This guide starts with a synthetic color check, then builds a release review around the real composition. It is a different question from whether a portrait crop retains the subject or whether the scene can be paused: the words need enough dependable contrast whichever frame, poster, or background the visitor sees.

Follow one sentence through the picture pipeline

The art team often approves a still exported from the source edit. The web team sees a responsive composition with an optimized movie, poster, overlay, and a title edited by someone else. Those two artifacts can differ even when they use the same artwork. A contrast check should follow the sentence the actual visitor reads, not assume that approval of a background picture approved all future uses.

Keep the copy and the image references separately identifiable. For an invented evening workshop, the campaign asset is Revision 7, the layout is Revision 3, and the small-print copy was lengthened after the first approval. If the long line now wraps into a pale area, a new concern has appeared without an image change. Record all relevant revisions when testing. A screenshot dated Tuesday but not tied to the build, crop, font and media can be difficult to reproduce Friday.

Treat motion as a set of states that can occur in normal use. The image might start as a poster, switch to video, freeze on a scrubbed frame, and go back to a fallback if the file cannot load. Position can vary independently while the learner rotates a device or increases type size. A state chart does not need hundreds of combinations at first. List the distinct ways the background under the sentence can change and tie each to a manageable representative case.

Some layouts deliberately fade both imagery and controls in an entrance. A reader arriving mid-transition can see small type while its opacity is still increasing. Measure or inspect the actual period during which the control is usable and its label expected to be understood. If the UI enables an action before its visual state is settled, a stable-looking final screenshot did not exercise that time. Consider eliminating the fade on essential text instead of promising to evaluate every animation instant on every device.

Name the text and its actual background

Start with the content that must always be readable. The museum runs this example with four foreground elements: the event title, opening time, venue address, and ticket action. They don't all occupy the same part of the media. The title sits near the top of the frame. Small address copy overlaps the doorway. The ticket action uses a solid fill, but its focus indicator can cross both the fill and the bright image. Checking a brand palette unconstrained by placement doesn't tell you how those combinations behave.

The W3C explanation of WCAG 2.2 minimum contrast describes a 4.5:1 ratio for normal text and 3:1 for qualifying large-scale text, along with exceptions. Don't treat any heading element as automatically large enough. Check its actual rendered size and weight against the applicable definition, especially on phones and where the title wraps. Separately review non-text contrast for qualifying essential graphics and visual information needed to identify controls or states. A text ratio is not proof that a control's boundary or focus state is discernible.

Don't assume that a theme preferring dark mode is proof the picture behind the title is dark. The Media Queries specification describes prefers-color-scheme as a preference for a light or dark color theme. It doesn't constrain the media file. Dark mode can deliver bright daylight footage and light mode can deliver a dim rehearsal photo. Keep theme state and media state separate in the review.

Write down what lies under each essential element. It could be the raw frame, a gradient over the frame, a tinted plate, a partially transparent button, a video poster, a generated fallback, or a loading color. Some pages use a different surface when media pauses. Treat that as an explicit state, not an implementation detail somebody can guess from the word “fallback.”

Measure a known pair before testing a whole clip

WCAG contrast calculations use relative luminance for the foreground and background. The ratio adds 0.05 to the lighter and darker luminances and divides the former by the latter. Relative luminance uses linearized sRGB channel values in the sRGB example. Computing contrast from the plain average of hexadecimal channels can produce a deceptively comfortable result. Use a correctly defined calculation or a reliable tool that specifies its inputs and assumptions.

Here's a fixture using solid, already resolved colors. A white title over the dark representative frame is strongly contrasted. The same white over the bright frame fails the illustrated normal-text minimum. A solid dark plate gives a stable sample pair. This fixture deliberately does not simulate CSS opacity, blend modes, HDR conversion, color profiles, anti-aliasing, or platform rendering.

def linearize(channel):
    if not 0 <= channel <= 255:
        raise ValueError("expected an 8-bit sRGB channel")
    s = channel / 255
    return s / 12.92 if s <= 0.04045 else ((s + 0.055) / 1.055) ** 2.4

def luminance(rgb):
    r, g, b = (linearize(c) for c in rgb)
    return 0.2126 * r + 0.7152 * g + 0.0722 * b

def contrast(fg, bg):
    low, high = sorted((luminance(fg), luminance(bg)))
    return (high + 0.05) / (low + 0.05)

white = (255, 255, 255)
black = (0, 0, 0)
hallway_shadow = (20, 24, 36)
open_daylight = (240, 233, 212)
plate = (24, 28, 42)

assert abs(contrast(white, black) - 21.0) < 1e-9
assert contrast(white, hallway_shadow) > 4.5
assert contrast(white, open_daylight) < 3.0
assert contrast(white, plate) > 4.5

for name, color in (("shadow", hallway_shadow),
                    ("daylight", open_daylight), ("plate", plate)):
    print(name, f"{contrast(white, color):.2f}:1")

Compare the unrounded ratio to the threshold. Rounding up a result near the cutoff is not a pass. A practical design also has a reason to aim for margin: minor font or rendering differences, compression artifacts, viewing conditions, and display settings can affect apparent ease of reading. Don't substitute a comfortable margin target for the actual requirements of your release; report both if you adopted an internal threshold.

An image behind text is rarely solid. If you inspect a screenshot, don't sample an anti-aliased pixel on the edge of a glyph and treat it as the foreground token. Start with the intended computed foreground, then inspect the background surface spatially behind relevant glyphs and at the same viewport and frame. A paired preview that hides the type while leaving all compositor layers in place can help locate the background; verify that hiding it didn't alter the layout or remove something like a text-shaped effect you meant to evaluate. For glass, gradients or blending, test the final composed surface rather than claiming a theoretical RGB pair from incomplete inputs.

Calculate the risk of a partly transparent surface

An overlay color is not its displayed background color when its opacity is less than one. Its final appearance depends on the scene underneath and on compositing as implemented on the target. A dark translucent scrim can produce a generous ratio over the workshop doorway when the door is closed and a much smaller one when the wall behind it turns white. You can reason about a bounded set of resolved colors without falsely claiming to have tested every frame.

For a synthetic resolved-color exercise, compare white text on RGB 40, 40, 40 with the same text on RGB 128, 128, 128. The former easily clears a 4.5-to-1 reference; the latter is about 3.95-to-1 and does not. These are hypothetical final backgrounds used to explain a contrast change, not results of any specific opacity percentage. If a designer says “it is black at fifty percent,” that isn't enough to identify the final backdrop, especially if other transparent layers, browser effects or color transformations are present.

In the same exercise, the column for foreground should change if the control's text becomes translucent on hover. Comparing its nominal white design token against the visible backdrop would give the wrong answer. Verify which opacity applies to which element. A background color with transparency is different from a parent opacity that fades the child text and its background together. Deliberately move the authority for contrast to an opaque inner plate or separate text when the chain of effects becomes hard to bound.

A useful frame review keeps the computed foreground, intended plate, and observed background as different entries. In a controlled solid state they may be straightforward to predict. For blended states, freeze an identified frame and layout, preserve the active layers, and inspect each location without the foreground where possible. Log the representation and limitations of any tool used. Avoid quoting ratios derived from one unqualified screenshot if a renderer or capture pipeline changed color encoding or coverage in ways you can't resolve.

Test how much certainty the design really needs. An optional decorative title layered over art might not be where visitors obtain essential facts. A ticket action or time must communicate regardless of the asset. A smaller, controlled region may carry the important information while the rest of the treatment remains visually expressive. Less dependence is often easier to verify than a more sophisticated dynamic overlay.

Compare three affordable design choices

An opaque text plate is the easiest surface to reason about for contrast. It can be compact and leave most of the footage alone. Use approved solid text and plate colors; don't apply opacity to a parent that also fades the type. Keep imagery from overflowing on top of the plate. In the museum example the address, hours, and action live on a small solid panel, and the video's strongest area remains visible beside it. A sufficiently clear plate solves the words without dimming the whole scene.

A translucent scrim can preserve a more atmospheric overlay. It also passes more of the media variability into the text background. A fixed scrim over a controlled library of calm clips can work if you understand the range and verify it. A demonstration on the chosen first frame is not the same as a bound over all frames or future uploaded clips. If a scrim is stronger at one side, wrapping or a responsive reposition can move small text out of the protected zone. Review it at the actual placement, not just at a point nearby.

Text outside the media is often the smallest dependable design. Put the title and important details on a normal document surface and use the clip as an adjacent visual. On phone layouts the clip can sit directly under a readable title instead of behind it. The museum might keep the overlay for the big heading on desktop but move vital timing and access information to an ordinary text area on narrow screens. This protects usability if images, video, animation or custom fonts are unavailable. It can also make translated text fit without asking a frame to provide more safe space than it has.

Be wary of depending on fast-changing adaptive color to make text legible. Automatically flipping the foreground from black to white on scene cuts introduces synchronization, palette, visual stability, and control-state questions. A short transition may produce two colors both readable against neither the old nor new background. If you genuinely need adaptive type, define how it behaves between sampled frames and across multiple elements. For most small publishing teams an opaque plate or separate text is less fragile.

Review the protected region as copy changes

A common failure isn't caused by a brighter pixel. The protected region ends exactly where the shorter caption ended. A translation wraps over a second line, a larger system font increases height, or an event location contains a long name. The new text lies partly outside the scrim. A screenshot that accounts for a bright moment but only with the shortest copy misses it. Define a region that flexes with the text or use a layout that doesn't depend on a fixed-size dark area.

In a fictional mobile layout, the address plate starts as twenty-six characters on one line. An amended accessible entrance note adds a second line and a separately linked door location. The address remains covered except for the link's focus ring extending several pixels beyond the plate. Two assertions now need checking: legibility of the address and appropriate visibility of the focus state at the plate boundary. Changing the text ratio cannot solve the edge condition.

Check positions after known responsive switches. A headline might jump from upper left to bottom center when the media changes from a wide clip to a portrait crop. Do not reuse a check made at the same height and width on the original export if the image section in view is different. Give test receipts enough information to reproduce placement: viewport, text scaling, document zoom if used, wrapper composition, selected crop and the actual media state. A stylesheet token is evidence about intended style, not which pixels ended up beneath it.

A compact visual fixture can turn some of these checks into safe predictable tests. Put a neutral light area behind the edges of a flexible plate and a dark area in its center. Add a long synthetic translation and push the layout to a supported narrow breakpoint. Inspect whether any required text or visual control information can escape. If the design expects a separate document layout there, verify it actually switches before a partial plate state occurs. A pressure test with simple blocks can reveal an implementation issue without replaying the whole artist's clip repeatedly.

Inspect where the text sits, not just which frame is brightest

Build a frame review around relevant locations. A whole-frame brightness histogram might rank a clip as mostly dark while the one highlight under the address is nearly white. The reverse can also be true: the brightest frame overall may have a naturally dark rectangle exactly where the type appears. The review should track the regions under the actual text, including each line after wrapping and the area around the action control.

Start with several witnesses: the poster, beginning and end of the clip, maximum-light events like the doorway opening, scene cuts, the transition or loop seam, and frames where visible objects move behind the copy. If the editor exports known markers, use them as a first set rather than the whole proof. Sampling by second alone can miss a short reflection or a frame the browser holds while buffering. A fixed-frame sample is useful for reproducing what you saw; it doesn't guarantee the unseen frames.

SurfaceWitnessWhat to inspect
Wide light themeOpening poster, bright doorway, seamHeadline, small address and hours
Narrow light themeSame witnesses with actual cropWrapped lines and top edge of tickets control
Narrow dark themePoster and bright moving frameSmall white text and focus state
Motion unavailableNo video, poster delayed or missingFallback color, address, primary action
User zoom or larger textRepresentative wrapped layoutType doesn't leave plate and control stays usable

An abstract frame review is not a replacement for a functioning page. Capture screenshots or a brief trace at the preview's actual route and keep the asset revision. Confirm your CSS target sizes, whether a banner or browser UI moved the layout, and if the device was zoomed or in a different orientation. When the sample fails, the next person needs to find the same composited area, not retest a convenient page that looks similar.

Treat posters and loading states like routes in the product

A media poster can be updated independently of the video, often by somebody selecting a good promotional frame. If poster copy relies on the same overlay as the motion, it deserves a recorded test of its own. Check the poster at multiple supported widths and whether a side crop changes which object sits behind the controls. A static first impression may need a fallback surface even if the chosen video has a reliably dark section.

An absent or slow poster shouldn't erase time and place. Use a deliberately named container surface, approved text colors, and legible controls while media is waiting or unavailable. In CSS implementations, a nested child may have an image background while its parent has a defined fallback. Test which surface appears when the child isn't painted, and what happens when the overlay disappears along with it. The protected text must not inherit a default for a background that no longer exists.

For an invented scenario, simulate three states on a supported preview: movie blocked but poster present; both image resources unavailable; and video paused on a representative bright crop. Check the address, fine print and focus-visible path in each state. A reload showing the slowest observed startup still doesn't prove the absolute longest timing, but demonstrates behavior with a controlled resource missing. Mark which tests were simulated rather than giving the impression you actually observed visitor devices under all network conditions.

If a reduced-motion preference changes the composition, inspect the result rather than assuming it uses the same poster and overlay. A design may swap the movie for still art with a lighter center. Another may turn off the whole element and change text alignment. The scope here is not whether the system honored a motion preference in every way. It's whether the essential copy stays readable in the actual static mode the product presents. Keep that distinction when handing a bug to another owner.

Check states that appear only during use

The museum header uses a focused link over the media at the far right. That link may look fine until the person tabs into it and its ring is nearly invisible against the pale doorway. A ticket control that darkens on hover may differ from its keyboard focus or active state. Style the combinations as deliberate surface pairs, and photograph or measure the states from a real navigation path. This is separate from checking the button text ratio in its default state.

Check loading and errors without making the layout disappear. Slow startup may leave a plain background before the poster decode. A broken poster may expose a default transparent or white surface. A video error could unexpectedly switch the element's rendering layers. Essential information should survive any of these states. Verify the hero at a width where the copy wraps and a height where a sticky toolbar can press it downward. The default background color of the container can be the simplest insurance if it was selected and tested deliberately.

Some browsers or high-contrast modes can force colors or change how backgrounds, links, and borders render. Preserve ordinary semantic links and use the system affordances rather than making a custom visual indicator the only feedback. A specifically composed dark frame isn't a substitute for a recognizable control. The release question is whether the viewer can find and use the action on the surfaces you're claiming to support.

Notice which colors touch when the controls are active

Some visual affordances cross more than one surface. A focus ring can start on an opaque plate and extend into moving imagery. A button border can touch a translucent edge whose contrast changes sharply between light poster and dark video. Don't reason from a single preferred backdrop for the entire shape. Observe which parts communicate the state and what they touch in the conditions you support.

Review using the means people actually use. A mouse-hover snapshot cannot stand in for keyboard focus, and a simulated pressed color may not match the touch experience. A component that changes its fill when active can be fine at rest and provide an inadequate contrast then. Check the actual sequence that produces the state: arrive on the page, reach the button with the expected input, activate where safe in a test fixture, and return. Record any states intentionally not tested.

A stable solution might place an inset state indicator against a known color or give the action more room within a controlled panel, while respecting supported visual and semantic behavior. Don't promise that a thicker outline alone meets every criterion. The named standard and applicable state determine what needs to be discernible. Similarly, a contrast-proof text plate does not prove sensible keyboard order or functionality. Label the scope honestly.

Keep labels as real text and actions as actual controls when appropriate. An ornamental outline in a baked movie frame has no interactive state. A “Tap here” sign in the video might appear to invite behavior that only a real control can supply. If a campaign crew wants that visual, keep it decorative and put the usable action in the interface with its own stable label and visible state. Separate the proof that a person can locate it from the proof that its colors meet a chosen threshold.

Run a single review that forces a design decision

Give the team a preview page that can freeze the clip at the actual doorway-opening frame. The first screenshot shows the current composition. The address is hard to read in the phone crop, and the foreground token itself is already white. Changing it to a different pale tone will not create a predictable background. Capture the area behind the address without the text so the team can distinguish a bright highlight from a low-opacity text style or an unexpected gradient stacking order.

Now make two controlled alternatives: one with an opaque plate that expands with the copy, and one that places the opening time and address under the clip. Leave the image and words identical. Reproduce the wide and narrow layouts, turn on longer labels, and compare the actual foreground-background combinations. The plate is likely less disruptive on the wide page. The separate document layout might be more reliable on phone when a long venue name wraps into three lines. Those outcomes can legitimately use different treatments rather than passing or failing a single design everywhere.

The review also needs a negative sample. Temporarily remove the plate token in the preview fixture while keeping the composition, then verify the check catches the bright-frame failure. This is useful when people intend to automate a screenshot check; a test that lets both the deliberate regression and the proposed fix pass is not measuring the right thing. For a fixed synthetic palette, log measurable ratios. For a real clip, label each observation by its selected frame and location instead of claiming the frame sampler covered every possible transition. Don't train the team to conflate a proof on clearly resolved solid tokens with an eyeballed test on a glass panel.

Create a set of frame witnesses you can grow

A useful review starts with likely stress moments and has a way to add discoveries. Select scenes containing bright objects behind the known text region, short uncontrolled transitions, a loop seam, media ending behavior, and scenes where the source art typically becomes lighter. If a fixture freezes a frame, preserve the actual timestamp or other reproducible identifier. Then review normal playback as available; a single frozen example tells you why something failed, not how broadly the issue occurs.

When a human watching finds a missed flash, record the witness without rerunning every frame blindly. Confirm whether the affected object overlaps a real text position under a supported crop. Compare fixes by replaying the same witness under the same setup. A small intended improvement should not be validated using a different time when the old design would already have passed. Keep one intentionally bad variant while testing the method; it should make the failure easy to reproduce.

A timeline plot of brightest pixels in a zone might help prioritize review if you can define an accurate zone and reliable capture pipeline. It won't tell you that a certain pixel lay behind a relevant part of the glyph in every affected layout. A broad-region witness can overstate some hazards while missing others outside the zone. Treat it as a screening method, not the complete verdict. For a controlled opaque plate, a known color pairing and geometry check can establish a different and stronger kind of evidence.

If the product manager wants a guarantee for arbitrary future videos, state the design condition that makes it supportable. For example, essential address and time remain on an opaque controlled backdrop independent of media; the footage beneath is decorative to their reading. Verify the color pair, coverage at supported layouts and interactive states separately. You can't derive universal future readability from ten sampled scenes in one day's footage. Choose a bounded layout that supports the claim or narrow the claim to observed material.

Use an export handoff that names its limits

Media and web releases can move independently. A contract between them can be small: here are the approved variations, poster, color-critical interface elements, intended protected regions, text revisions, and retest triggers. If the critical type is never placed on the picture, say so. If a target deliberately uses a particular dark region in the clip, explain that substituting a new source invalidates that assumption. The recommendation should travel with the composition, not rely on a specific reviewer remembering which frame caused trouble.

Suppose the production editor sends a new clip only to correct a logo at the end. A developer sees an updated asset and asks which QA has to be repeated. A stable plate color pair may be unchanged, but the new codec or poster can affect the fallback; the new end frame can affect exposed labels; changes to frame duration can alter transition states. A targeted retest follows the affected paths. Do not invalidate genuinely independent checks merely because a build ID changed, and don't claim a check survives without knowing what it depended on.

A good evidence record includes version, asset and poster, foreground element, layout context, identifier for the selected frame, actual background treatment, target criterion, numerical result if supported, and unresolved conditions. A concise receipt in a task ticket can capture all of that for one blocking defect. Preserve a benign reference with made-up content where needed so employees needn't share restricted footage just to reproduce a blur or faded edge.

Use an explicit decision vocabulary. “Passes tested stable pair” describes a calculation. “No failure seen in sampled bright-frame review” describes observation. “Not inspected in high-contrast setting” describes missing evidence. Mixing all three under a green badge makes the claim hard to repeat and hard to correct. The aim is not to make the reviewer drown in caveats. It is to make an appealing moving design robust wherever essential words and controls have to remain recognizable.

Make the alternate composition a fair comparison

A design review sometimes pits a fully polished image overlay against a rushed card with none of the typography, spacing or color details that made the original attractive. The comparison appears to endorse the overlay, but it changed more than the vulnerable surface. Preserve the copy, image, scale and approximate hierarchy where practical, and change only the source of the backdrop. Compare the alternatives on the hard frame before asking colleagues which one feels more compelling.

For the fictional workshop page, keep the same still crop and headline in both candidates. Candidate A places the address and opening hours inside an expanding dark panel. Candidate B puts those two lines on a stable page surface below the art. A bright stripe crosses the center of the poster. Test the long addresses again on a narrow screen and verify logical reading and focus order after the move. A design that protects the address but puts the ticket action before the venue heading may have solved only one of the problems.

Review visually against the newest content, but avoid counting personal reactions as proof of ratio. One useful internal worksheet includes measured pairs where available, discomfort or delayed reading observed by reviewers, control visibility, layout overlap, and whether qualified users can still locate the important action. Those are complementary checks. It takes only one source with the right criterion and enough reproducible evidence to justify a contrast correction; it shouldn't take unanimous dislike of the art.

Some test outcomes warrant simplifying the requirement itself. If a clip contains subtitles or animated lettering that constitute actual information, moving the interface title off the video doesn't automatically solve that content. Record what belongs to the interface and what is part of the asset. If presentation differs because a third-party embed controls captions, check its supported interface and don't make claims about a color treatment you don't control. A page's title and action can still be covered by a good local release decision.

Separate perceptual stress from numeric compliance

A valid ratio is an important technical result for a defined pair. It is not a guarantee that a sentence is effortless to read over textured moving imagery with lots of detail nearby. Position, motion, weight, letters, spacing and surrounding complexity can affect usability. If a sample meets the relevant threshold but the practical reading exercise repeatedly fails, making the text more stable can still be the right decision.

One synthetic test can ask a participant to find a specific opening time and an entry location without prompting them where to look. Run the same task with the same content against the alternate layout and document what happened. This is a practical usability probe, not a replacement for WCAG evaluation or a statistically representative trial. It can explain why a reliable plate with enough spacing earns its place even when a carefully specified original pair passed a narrow calculation.

Write down the threshold and the proof

The most credible status note identifies the actual values and the limitation: “On the narrow preview, the address overlays a solid #181C2A plate and uses white text. The specified pair passed the arithmetic check. We verified the plate covers the wrapped address on the poster and bright-frame screenshots. The separate focus ring was checked visually in the keyboard path; browser color override was not tested.” That is more useful than “dark mode passes” because the next designer can tell which result survives a new clip.

Require another check if someone changes text color, width, overlay position, video treatment, or the foreground tokens used in each theme. If copy alone changes, test whether new line breaks leave the protected region. If the video changes but the fully opaque plate and text arrangement are unchanged, the plate's pair remains stable; the cropped story, non-plate foreground, and any controls that overlap the frame may still need review. Use the previous receipt to narrow testing rather than starting every release from zero.

It also helps to leave an approved fallback composition in the design file. Treat it as a first-class state. An empty video error box or a poster with title baked into it can undermine accessibility even if the regular moving view is beautiful. The best fallback has the same useful title and action in document text, a stable surface, and no dependence on the visual asset conveying the basic facts.

The payoff is visible at the bright frame. The door opens, the scene still feels alive, and the person looking for the address doesn't have to wait for a better moment to read it. You can keep the moving picture and the design's mood. Just choose who gets responsibility for text contrast: a controlled surface the team can verify, or a video frame the visitor can't.

Frequently asked
What is Contrast Over Moving Pictures: Prove the Text Survives the Bright Frame about?
A museum page promotes an evening event with a clip of light entering a workshop. It begins with a dim central doorway. White text looks unusually sharp on…
What should you know about follow one sentence through the picture pipeline?
The art team often approves a still exported from the source edit. The web team sees a responsive composition with an optimized movie, poster, overlay, and a title edited by someone else. Those two artifacts can differ even when they use the same artwork. A contrast check should follow the sentence the actual visitor…
What should you know about name the text and its actual background?
Start with the content that must always be readable. The museum runs this example with four foreground elements: the event title, opening time, venue address, and ticket action. They don't all occupy the same part of the media. The title sits near the top of the frame. Small address copy overlaps the doorway. The…
What should you know about measure a known pair before testing a whole clip?
WCAG contrast calculations use relative luminance for the foreground and background. The ratio adds 0.05 to the lighter and darker luminances and divides the former by the latter. Relative luminance uses linearized sRGB channel values in the sRGB example. Computing contrast from the plain average of hexadecimal…
What should you know about calculate the risk of a partly transparent surface?
An overlay color is not its displayed background color when its opacity is less than one. Its final appearance depends on the scene underneath and on compositing as implemented on the target. A dark translucent scrim can produce a generous ratio over the workshop doorway when the door is closed and a much smaller one…
References & sources
  1. Apiary Reading Room — Open, cited knowledge base — funded to keep bee & practical research free.
From the Apiary Reading Room. Opinion & editorial — not financial advice. We don't overclaim.
More from the Reading Room