Devicedial
32 free browser tests
Every device test on this site
Grouped by what you are holding: the things that carry your voice and face, the things you type and click with, the things you play on and look at, the tones you send through them โ and the sheets that keep the readings once you have them.
Mic, sound & camera
Camera test
Two numbers the camera reports about itself, and what each one is costing you.
Echo test
Five seconds in, five seconds back โ and the room comes with it.
Headphone test
Two of the three faults are obvious. The third only shows when you switch between them.
Mic test
Speech should peak inside a band most people have never been told about.
Speaker test
Left, right, or both โ and seven frequencies that go missing for different reasons.
Webcam mirror
A screen makes a better light than a bathroom mirror ever gets. The lens is the part that lies.
Keyboard, mouse & reflex
Aim trainer
Speed and accuracy trade against each other, and target size moves the score more than either.
CPS test
Most people are slower than they assume โ and the part they blame on the mouse usually is not the mouse.
Double click test
One press, two clicks. There is a number that separates a worn switch from a fast finger.
Keyboard tester
A cheap board gives up at two keys; the usual USB report carries six. Where does yours stop?
Mouse tester
Five buttons, two wheel axes, and one number the browser is not allowed to know.
Mouse polling rate test
1 ms between reports is 1000 Hz. Whether your browser can even see that is the interesting part.
Reaction time test
Part of every reaction time on the internet is the monitor. This one says how much of yours.
Typing speed test
Speed without accuracy is slower than it looks. See both numbers from one 60-second run.
Controller & screen
Dead pixel test
Before you count the spots on your screen, there is one step almost everybody skips.
Gamepad tester
Every button, every axis, live โ with the number behind the dot, not just the dot.
Refresh rate test
Your monitor is set to 144 Hz. Whether it delivers 144 is a different question.
Screen resolution
The browser says 1707 by 960. The box said 2560 by 1440. Both are telling the truth.
Stick drift test
A dot that wanders is not a verdict. Ten seconds of measurement is.
Touch screen test
Ten fingers, one panel. Most screens stop counting long before you run out.
Tones & hearing
Audio test
Two numbers your operating system knows and never shows you.
Bass test
A speaker that cannot play 40 Hz can still make you think it did.
Frequency test
Both ends of your range in one pass, plus everything odd in between.
Hearing test
The result belongs to four things at once, and only one of them is your ears.
Surround sound test
The number of channels your output accepted, before any speaker is blamed.
Tone generator
A frequency you name, at a level you can quote back.
Test plans & logs Pro
Gear test schedule
Set the interval once and get the next twelve dates for the tests that catch a fault early โ with an email a few days before each one, if you want it.
Headset and mic comparison
Two or three microphones side by side with the peak and the background noise you measured, and the gap between them in decibels for each one.
Input lag budget
Every wait between your hand and the screen on one sheet โ polling interval, frame time, panel response, render queue โ added into one worst-case figure in ms.
New machine test plan
A target for every check, your own reading beside it, and a count of how many the machine actually met โ run it while the return window is still open.
Stream check sheet
Every source you are sending โ camera, screen share, audio โ turned into megabits a second, added up, with headroom and the size of a two-hour stream.
Warranty claim log
A dated row for every reading, the limit it should have stayed inside, and a running count of how many went past it โ evidence instead of a description.
What can you actually test in a browser?
Devicedial tests six things a browser can measure honestly: the microphone, the camera, the keyboard, the mouse, a connected controller and the screen. The web platform hands a page the live audio and video streams, every key and pointer event, the gamepad axes, and the timing of each frame it draws โ which is enough to produce a real figure for level, resolution, rollover, polling rate, stick drift and refresh rate.
What it deliberately does not do is guess. Mouse DPI, panel sampling rate and the hardware behind a driver are invisible to any web page, so those tests say so rather than estimate.
The rule across the site is that a test ends in a number on a named scale โ dBFS, hertz, milliseconds, pixels โ and a sentence explaining what that number means. A green tick tells you something is connected. It does not tell you that a microphone peaking at โ34 dBFS will sound like a whisper to everyone in the call.
A number first, then what the number means
Most device testers answer yes or no. A green tick tells you the browser can see your microphone; it does not tell you that you are recording 14 dB too quietly, that your mouse is reporting at 500 Hz instead of the 1,000 you paid for, or that two of your clicks were one press. Every test here ends in a figure on a named scale, and a sentence about where that figure sits.
frame time in ms = 1000 ÷ refresh rate in Hz
The scale each test reads, and where the line sits on it
- โ20 dBFSbelow this a mic sounds thin and hissy
- 16.7 msone frame at 60 Hz
- 1,000 Hza gaming mouse reports 1,000ร a second
- 50 mstwo clicks closer than this is switch chatter
- 5 charsone โwordโ in every WPM score, space included
- ยฑ1.00the whole range a stick axis can report
- In range โ nothing to fix
- Borderline โ worth a setting change
- Out of range โ hardware or driver
Something in front of you is misbehaving and you want to know whether it is the hardware, the driver or the app. That is one question, and it has one honest answer: a measurement. Not “we detected your microphone” โ a level, a rate, a count of the things that answered.
Why a number beats a green tick
A tick means the browser found a device. It survives a microphone recording 14 dB too quietly, a mouse reporting at a fifth of its advertised rate, a monitor delivering 60 frames while its box says 144, and a switch that sends two clicks for one press. Every one of those is a working device by the tick’s definition and a broken one by yours.
So each page here prints the figure and the scale it sits on: dBFS for level, hertz for rates, milliseconds for time, a count for buttons and axes. Then it prints the sentence that turns the figure into a decision โ in range, borderline, or a hardware problem. The figure is what you can quote to a support desk; the sentence is what you actually wanted.
The traps the units set
Half the confusion in this niche is unit confusion, and it is always the same handful. The dBFS scale runs downwards from a ceiling of zero, so โ6 is louder than โ20 and neither is silence. Hertz and milliseconds move in opposite directions: more hertz is better, more milliseconds is worse, and one frame at 60 Hz is 16.7 ms whether you like it or not.
Resolution has two answers at once, because a CSS pixel is not a physical one โ on a display with a device pixel ratio of 2 the browser reports half the width the panel actually has, so both numbers belong on screen together. And DPI is the one figure a browser genuinely cannot read; a page that prints your mouse DPI is guessing, and this one says so rather than inventing it.
What the rest of the site covers
32 tests, grouped by what you have in your hands. The sound and camera group covers the microphone, your speakers and headphones channel by channel, echo, and the webcam with its live resolution and frame rate. The keyboard, mouse and reflex group covers key rollover and ghosting, typing speed, buttons and wheel, double-click chatter, polling rate, reaction time, click speed and aim. The controller and screen group covers gamepad buttons and axes, stick drift, dead and stuck pixels, refresh rate, resolution, and touch input. The tones and hearing group goes the other way round โ the browser plays, you listen: a tone at any frequency you name, the top and bottom of what this system can reproduce, and how many channels it really has.
The test plans and logs are for after the measuring. A single reading answers today’s question; a column of them with dates is what settles a warranty claim, tells you which of two headsets is actually quieter, or adds one link at a time until you can see where your input lag comes from.
None of them needs a download, an account or a network connection while it measures. If a test cannot see something, it says so โ that is the whole difference between this site and the eight identical ones above it in the results.
Frequently asked questions
Do I need to install anything?
No. Every test runs inside the browser tab using web APIs that ship with Chrome, Edge, Firefox and Safari. There is nothing to download, no extension, and no account. The microphone, camera and controller tests ask for a permission because the browser requires one, and you can withdraw it from the padlock icon in the address bar at any time.
Is anything uploaded or recorded?
No. Audio and video streams are handed to the page by the browser, measured frame by frame, and discarded โ they are never written to a file or sent anywhere. The echo test is the only tool that holds audio at all, and it keeps five seconds in memory to play back before throwing it away. No test on this site makes a network request while it is measuring.
Which tests need real hardware?
Keyboard, mouse, polling rate and double click need a physical keyboard or mouse โ on a phone they would measure the on-screen keyboard, which is a different thing, so they say so. Controller tests need a gamepad paired over USB or Bluetooth; the Gamepad API is the same on phones and desktops. Microphone, camera, touch, screen and reflex tests work on any device.
Does this page record or upload my voice?
No. The audio never leaves the tab: the browser hands the microphone stream straight to an analyser that measures amplitude, the numbers are drawn on screen, and the samples are discarded frame by frame. There is no recording, no server call and no account. Close the tab and the only trace left is the permission you granted the site, which you can revoke in the padlock menu in the address bar.
Why does the browser ask for permission every time?
Because you are answering โAllow this timeโ rather than โAllow while visitingโ. Browsers deliberately make one-off consent the easy click for camera and microphone access. Choose the persistent option in the prompt, or open the padlock icon next to the address and set Microphone to Allow, and the question stops. On a shared machine the one-off answer is the better habit even though it is the more annoying one.
Do these tests work on a phone?
Most of them do, and the ones that cannot are honest about it. Microphone, camera, touch, screen resolution and reaction time all work on a phone. Keyboard, mouse and polling-rate tests need a physical keyboard or mouse, so they either report what a touch device can report or say plainly that the reading would be meaningless. Controller tests work on a phone with a paired gamepad, because the Gamepad API is the same one desktop browsers use.
One gear check a month
MonthlyOne thing a month worth measuring before it fails properly โ a mic that reset itself, a stick that drifts, a switch sending two clicks. Two sentences and the test.
How the test works
AnalyserNode on the live stream and reported in dBFS, decibels relative to full scale, where 0 is
the maximum a digital sample can represent and every real signal is a negative number. Frame timing
โ refresh rate, frame time, and the millisecond figures in the reaction tests โ comes from
timestamps supplied by requestAnimationFrame, so what the browser reports is the rate of frames
it actually delivered, which can be lower than the rate your monitor is set to. Mouse polling rate
is measured by counting pointer events per second, capped by what the browser is willing to
deliver; typing speed uses the standard definition of one word as five characters including the
space; click chatter is two falling edges of the same button closer together than the threshold on
the page. Stick drift is read from the normalised โ1 to 1 axis values of the Gamepad API, which is
a ratio rather than an angle, so the drift score is our own definition and the page says how it is
worked out. Where a browser genuinely cannot see a value โ mouse DPI is the clearest case โ the
test says so instead of estimating.Sources: