Webcam Test

A webcam can show a live picture while a bright background hides your face or uneven frame delivery makes movement look jerky. Before an interview, video meeting or proctored exam, a useful camera check considers access, visible detail and motion together. Seeing a picture confirms that video is reaching the browser; judging whether that picture is usable takes a closer look.

Access problems and picture problems have different causes. A website needs camera permission, and the operating system may separately restrict the browser. A closed privacy shutter, disconnected camera or another application using the device can also prevent a useful picture. When several cameras are available, successful access may still open a different camera from the one intended.

Three terms help distinguish the information in a camera check.

Resolution

The width and height of each video image in pixels. A 1280 × 720 picture contains more pixels than a 640 × 480 picture, but its dimensions alone do not establish focus or lighting quality.

Frame rate

The number of video frames delivered per second, usually written as fps. A browser-reported setting describes the configured stream rate; an observed rate counts frames over a period of time.

Frame timing

How evenly those frames arrive. An average rate can conceal pauses followed by bursts. Variation in the spacing between observed frames is called jitter.

The browser negotiates camera settings with the device and its software. A requested size can be a preference rather than a guarantee, and video may be cropped or scaled before presentation. Even a successful high-resolution request establishes the output available through that browser, not the camera sensor's maximum resolution or every native capture mode.

Lighting changes both appearance and performance. In low light, a camera may lower its frame rate. Automatic exposure and focus also change the picture as the scene changes. A bright window behind someone can dominate the image while their face remains dark, so average brightness is not enough to judge visibility. Diffuse light facing the subject and a stable camera position make comparisons more useful.

Brightness, contrast and edge detail describe different aspects of the image. Contrast measures how widely brightness values vary; sharpness estimates respond to local edges. A plain wall can have little contrast or edge detail even when it is in focus, while noise or a textured background can raise these values. Such checks provide clues about the scene and camera output, not calibrated readings of room illumination or optical focus.

A local preview also leaves part of a video call untested. Microphone access, internet upload performance and the receiving service affect the eventual experience. A smooth self-preview cannot establish what another participant receives, and an exam provider may impose its own requirements. Camera checks are most useful alongside a visual inspection and the intended service's own system check.

How to Use This Tool:

Use the camera position and lighting planned for your call. For an ordinary check, leave Preferred size at Automatic size and Measurement window at Standard (10 seconds).

1. Click Start camera and allow camera access if the browser asks. A live picture should appear before the measurement can finish.
   If access fails, open Troubleshooting for the actual error. Permission failures can involve site settings, operating-system privacy settings or managed-device restrictions. If the camera cannot be read, close other camera apps and check its connection and shutter. Open the HTTPS page directly in a current browser, correct the reported problem, then click Start camera again.
2. Select the intended device under Camera after access succeeds, then click Switch camera if needed. Refresh cameras updates the list after connecting another device; selecting a name alone does not switch the picture.
3. Choose different settings in Advanced only when needed. Preferred sizes are 640 × 480, 1280 × 720 and 1920 × 1080 pixels; the shorter measurement option is 5 seconds. Changes to size, duration or resizing apply on the next start, restart or camera switch. Click Restart test to apply changes to an active test.
4. Keep the tab visible until Measurement complete appears. Hiding the tab discards the measurement and starts a fresh window when you return.
   Completion freezes the measurements while the preview remains live. Use Restart test after changing lighting, position or other conditions you want to measure again.
5. Review Readiness criteria alongside your face and movement in the preview. Resolve relevant failures or warnings, then repeat the check. If you want a record of the completed result, use Copy report or Download report before stopping.
6. Click Stop when finished to release this tool's camera access. Then run the interview or exam provider's own check with the camera available to it.

Optional picture and size comparisons

Mirror preview changes only the displayed picture; sampled pixels and results stay unchanged. Resize preference can leave resizing to the browser, prefer no browser resizing, or allow cropping and scaling when supported. Camera requests prefer 30 fps, including with Automatic size, but the actual reported and observed rates can differ.

While the camera is running, Probe common resolutions in Advanced tries six sizes and briefly changes the preview. Read the comparisons in Resolution probe, then wait for a fresh measurement after the original size is restored. If restoration fails, capture stops; choose Automatic size and start again. Keep the tab visible during probing, because hiding it cancels the probe and stops capture.

Interpreting Results:

Camera readiness combines eight checks using this tool's stated defaults. This is a general check, not certification for any exam platform. The individual criteria explain a result more usefully than the summary alone.

• Ready means the measurement finished with an active camera and all eight criteria passing.
• Ready with warnings means the window finished with no failing criterion, but at least one check has a warning or remains untested. This can include approximate timing, so it does not always establish a verified minimum frame rate.
• Not ready means at least one criterion failed. Causes include insufficient delivered resolution, a verified frame rate below the minimum, or no active camera after an attempted test. Deliberately pressing Stop also produces this result; it is not a diagnosis of broken hardware.

Not tested is the initial result, and Measuring indicates collection is still underway without a failing criterion. A failure can appear before the window finishes. Treat Not measured as missing evidence, not a measured zero or proof that no frames were dropped.

If frame delivery is slow or uneven, compare another run with better light, fewer competing applications or a lower preferred resolution. Keep the camera, scene and measurement duration consistent when comparing results. A high average frame rate with poor Timing coverage cannot establish evenly spaced frames.

For exposure, contrast or sharpness warnings, inspect the picture before changing settings just to raise a score. Luma distribution shows how sampled pixels are spread from dark to bright. A bright background or dark clothing can influence whole-image statistics, and the central sharpness sample does not locate your face. Check whether facial detail and movement remain clear in the actual setup.

Technical Details:

Video delivery has both a count and a time component. Browser observations describe frames presented for display and the spacing of those observations; they do not directly measure the sensor's capture speed. Image statistics describe a small sampled picture taken during the same test.

Formula Core

Displayed rate and average interval

Let N be the increase in displayed-frame count between the first and last observations, and t their elapsed time in seconds. The measured rate f is in fps, and the average interval I is in milliseconds.

The preferred observations use browser frame-presentation counts and timestamps. If those are unavailable, browser playback counters provide a fallback. Their total includes displayed and dropped frames, so cumulative drops are subtracted from cumulative totals before calculating the displayed-frame increase. If only changes in video playback time are available, the result is labelled Approximate update rate and cannot verify individual frame delivery.

The standard window requires at least 10 seconds between observations, or 5 seconds when selected. Permission and startup time are excluded. During collection, no first frame within 5 seconds or no further displayed frame for 3 seconds ends an incomplete attempt and stops capture. A completed result describes that observation window; it does not continuously monitor later delivery.

Timing coverage and jitter

An observation counts as an individually timed interval only when the displayed-frame count advances by exactly one. Larger jumps still contribute to the average rate, but their unseen intervals are not assumed to have equal lengths. For k individually timed intervals, d_i is each duration in milliseconds and d̄ is their mean.

C is a percentage and J is jitter in milliseconds. Jitter requires at least two individually timed intervals and is unavailable with the approximate playback-time method. Its mean is calculated from that interval subset, which can differ from the whole-window average interval I. Browser scheduling can delay observations, so jitter does not isolate a camera fault or measure network latency.

Late and dropped-frame estimate

A shortfall against the reported rate and browser-counted drops may describe overlapping missing frames. The estimate takes the larger count rather than adding them. Let r be the reported fps, D the increase in browser-counted drops, E the expected count, S the pacing shortfall and M the combined estimate.

P is the late/dropped percentage. Expected frames are rounded to the nearest whole frame, with a half-frame rounded upward. When a dropped-frame counter is unavailable, only the pacing shortfall contributes. When no rate is reported, only browser-counted drops can support the estimate, using displayed plus dropped frames as the denominator. If neither reference is available, no estimate is shown. These comparisons do not prove that the sensor lost that many frames.

For example, a reported 30 fps over 10 seconds implies 300 expected frames. With 270 displayed frames and 20 browser-counted drops, the measured rate is 27 fps and the pacing shortfall is 30 frames. The combined estimate is 30 frames, not 50, giving 10% against a denominator of 300. That is exactly the allowed late/dropped limit; timing coverage, jitter and the other criteria still need to pass.

Relative brightness, contrast and clipping

Image analysis uses a sampled copy limited to 160 pixels on its longest side. The latest available sample supplies the image results, rather than an average of every frame in the window. For red, green and blue channel values R, G and B on a 0 to 255 scale, relative luma Y is calculated directly from those values.

For n sampled pixels, mean luma μ describes average brightness and population standard deviation σ describes relative contrast. No conversion to physical illumination is made.

Highlights are pixels with Y ≥245; shadows have Y ≤10. If n_h and n_s count those pixels, their percentages p_h and p_s are:

The histogram counts pixels in sixteen bins of width 16 across the 0 to 255 range. Values below 16 enter the first bin, values from 16 to less than 32 enter the next, and so on. Every sampled pixel contributes to exactly one bin. Highlight and shadow percentages use their exact cutoffs rather than entire end bins. With no usable image sample, image results remain untested and no distribution is plotted.

Centre sharpness

A four-neighbour Laplacian compares a pixel's luma with the pixels immediately above, below, left and right. The sharpness score V is the population variance of these responses q across the central half of the sampled width and central half of its height, excluding outer-border pixels. This covers roughly the central quarter of the image area.

Here m is the number of usable central responses and q̄ is their mean. Texture, motion, noise and autofocus all affect V. The fixed sampling size makes the calculation consistent, but the score is not a calibrated focus measurement or a direct comparison of sensor resolution.

Readiness rules and boundaries

These cutoffs are this tool's general heuristics, not camera standards or exam-provider requirements. Equality belongs to the passing side of every numerical cutoff. Exposure, contrast and sharpness checks require an active camera and usable image evidence; frame-rate and timing checks require a completed window.

Timing reliability also needs a reported frame rate, at least two individually timed intervals and a frame-based method rather than approximate playback-time updates. Missing picture evidence produces Not tested criteria, which prevents an unqualified Ready result. The displayed numbers use up to two decimal places; comparisons use unrounded values, so a value displayed at a cutoff can still lie just outside it.

Requested, reported and delivered sizes

A preferred size is a target for browser negotiation. Reported dimensions describe the resulting stream settings; delivered dimensions describe the video image available for presentation. Size comparisons use the long and short sides, so 480 × 640 and 640 × 480 are equivalent for readiness and matching.

The optional probe makes exact requests for 320 × 240, 640 × 480, 1280 × 720, 1920 × 1080, 2560 × 1440 and 3840 × 2160 pixels, reversing width and height for a portrait stream. A successful request is checked against both reported and delivered dimensions; a rejection records the failure. Cropping, scaling and camera-driver processing limit what can be inferred even when dimensions match. These six attempts describe the current browser and camera combination, not an exhaustive list of native sensor modes.

Privacy Notes:

Camera frames are analysed in your browser. This test does not upload frames, device identifiers or measurement results, record video clips, or request microphone access.

• The camera stays off until you explicitly start it. The browser may reuse permission you previously granted; starting a new test does not necessarily display another permission prompt.
• Stop releases this tool's capture and clears sampled image data, including the histogram. Timing, size and probe evidence can remain visible in the current page. Switching cameras releases the previous capture before opening the selected camera. Closing or leaving the page also releases capture. Merely hiding an ordinary test tab does not stop its camera stream.
• Frames, camera identifiers and results are not saved in persistent tool storage or included in the page URL. Only size, duration, mirroring and resizing preferences can be carried in the URL; restoring them never starts the camera.
• Reports and downloads are created locally when requested and contain no camera frames or device identifiers. Chart images contain the brightness histogram, not a webcam photograph. Copies you save or place on the clipboard remain under your control.

References:

• Media Capture and Streams, W3C Candidate Recommendation Draft, 9 October 2025.
• Media Playback Quality, W3C.
• Video Frame Callback, Web Platform Incubator Community Group Draft Report, 2 August 2024.
• Use your camera and microphone in Chrome, Google Chrome Help.
• Troubleshoot camera issues in a meeting, Google Meet Help.
• Troubleshoot video and audio quality in a meeting, Google Meet Help.