Monitor Ghosting Test
Sweep objects across five backgrounds to see ghosting, motion blur and overdrive overshoot — then tune your monitor's OD setting. Runs entirely in your browser.
In Solid blocks every row carries the same four objects at the same instant: a white disc, a black disc, a mid-grey bar and a chequered block — same column, five backgrounds, so you can compare one transition against another at a glance. On the black row the black disc has nothing to transition against and vanishes; that is correct, not a bug. The discs are drawn antialiased so their outlines are soft by design. The bar, the chequered block and the contrast-edge rectangle have hard, pixel-exact edges — judge sharpness on those.
Each bar is the gap between two browser frames; a full-height bar is twice your baseline interval. This is the browser's own timing, not a measurement of the panel.
Keyboard: click the strip, or tab to it, then Space pauses, ← → change speed, M changes pattern and Esc leaves full screen.
Timing your display's frame interval before anything starts moving — this takes a second or two.
Motion is paused because your system asks for reduced motion. This test is motion, so nothing starts until you press Start.
Your browser refused real full screen, so this is a full-window overlay instead — the browser's own bars still cover part of the panel.
This is a visual test. It shows you the artefact; it does not put a millisecond number on your panel, because measuring grey-to-grey honestly needs a photodiode against the glass rather than a browser. The four figures above are the ones a browser can measure: how fast it is being handed frames, and how far the object travels between them.
Everything runs locally: the test itself records, captures and uploads nothing, and sends nothing to a server. (The site's shared script may still prefetch a page you hover over — that is the site, not the test.) For the cleanest read use full screen — in a window the site's own film-grain texture is composited over the strip.
This test needs the HTML canvas API.
Your browser did not give the page a 2D drawing surface, which usually means canvas is disabled by a privacy extension, a strict enterprise policy, or a very old browser. Current Chrome, Edge, Firefox, Safari and Opera all support it. Enable canvas or open this page in another browser and the test will run.
How to run it
- Close everything heavy first. Other tabs, a game in the background, OBS, a video call. The frame-rate readout under the strip has to sit at your monitor's refresh rate, otherwise you are testing your PC and blaming your panel.
- Go full screen. In a window the strip is composited with the rest of the desktop and the site's own background texture sits over it. Full screen removes both.
- Track the object with your eyes. Follow it, do not stare at a fixed point. Smear is what the object looks like while your eye is moving with it; staring at one spot shows you sample-and-hold blur instead, and every non-strobing monitor on earth fails that.
- Start at 960 px/s. If it looks clean, step up to 1920. If it is already a mess, drop to 480. You want the speed where the difference between two overdrive settings is obvious — not the speed where everything is broken.
- Read the rows top to bottom. The black and dark-grey rows expose slow dark transitions, which is where VA panels fall apart. The light rows expose overshoot, which is far easier to see against a bright background.
- Change patterns. Solid blocks for general smear, Contrast edge for overshoot halos, Chequerboard for loss of fine detail, Moving text for the practical question: can you still read a name plate while you strafe?
What each artefact looks like
- A soft tail behind the object, roughly the object's own colour — slow grey-to-grey response. This is classic ghosting.
- A hard-edged halo behind the object — brighter than the background on the dark rows, darker than it on the light rows — overdrive overshoot, also called inverse ghosting. It is an edge, not a fade. Turn the overdrive setting down one step.
- A murky dark blur that only shows on the black and dark-grey rows — VA black smear. Dark-to-dark transitions are the slowest ones on a VA panel, sometimes by a factor of five.
- Crisp doubled or tripled copies of the object — not the pixels. Either your frame rate is not matching the refresh rate, or a strobing backlight is running out of sync with it.
- Even, regular stepping right across the screen — sample-and-hold plus a speed too high for your refresh rate. Check the css px / frame figure; above about 32 the stepping dominates and no monitor setting will change it.
Tuning overdrive, step by step
- Set the test to Solid blocks at 960 px/s and go full screen.
- Check the frame-rate readout matches your refresh rate and the graph has no red bars. Fix that first if it does not — a dropped frame looks exactly like a smear, and you will end up tuning against a fault that is not in the panel.
- Find the overdrive control in your monitor's on-screen menu. It hides under a different name on almost every brand: Overdrive, OD, Response Time, TraceFree (ASUS), Rampage Response (MSI), Smart Response (Samsung), AMA (BenQ), Overshoot (some LG). Set it to Off or its lowest step.
- Look at the black row. At the lowest setting you should see a soft tail dragging behind the white disc. That tail is the panel's native response time with no help at all.
- Raise the setting one step. The tail gets shorter. Go back to the screen and look again before touching anything else.
- Repeat one step at a time. Stop at the last step before a bright halo appears behind the object.
- Confirm on the dark-grey row as well. VA panels often want a different step from the one the black row suggests, because dark-to-dark is a slower transition than black-to-white.
- If your monitor does not advertise variable overdrive, repeat the whole check with your desktop set to 60 Hz. A level that is perfect at 144 Hz frequently overshoots badly at 60, and that is the refresh rate a lot of desktop and video content actually runs at.
Why you stop one step early. The halo is overshoot: the driver pushed the pixel past its target colour and it now has to fall back. Your eye is far more sensitive to a hard, bright edge than to a soft dark tail, so the "faster" setting that produces halos usually looks worse in a real game than the slower one it replaced. Overshoot is also permanent damage to the image in a way smear is not — smear fades, a halo draws a bright outline around every moving object on screen.
What is actually being measured
Three numbers get mixed together constantly, and they are independent of each other.
- Refresh rate (Hz) is how often the monitor is handed a new image. 144 Hz means a new frame every 6.94 ms. It says nothing whatsoever about how fast a pixel can change.
- Response time (GtG) is how long a pixel takes to travel from one shade to another. If a panel needs 12 ms to finish a dark-to-bright transition and a new frame lands every 6.94 ms, the pixel is permanently mid-transition and everything in motion smears. That is what the strip above shows you.
- Input lag is the delay between your mouse moving and the light on the screen changing. It is dominated by the monitor's internal processing and the rest of the chain. A monitor can have fast pixels and awful lag, or the reverse — this test cannot see it at all.
Response time is a matrix, not a number
A panel does not have a response time. It has one for every pair of shades, and reviewers publish them as a grid for exactly that reason. The five rows here exist because the transition is what matters: 0 → 255 on the black row is usually near the panel's best case, while 0 → 50 on the dark-grey row is where a cheap VA panel can be five to ten times slower and produce the murky trail people call black smear. IPS and TN panels are far flatter across the matrix, which is why they look consistent even when their headline number is not the lowest.
The chequerboard pattern makes this explicit — each row scrolls a different transition pair, printed in the row label, so you can see which part of the matrix your panel is bad at rather than guessing.
What overdrive actually does
A liquid crystal cell moves faster when you apply a bigger voltage difference. Overdrive exploits that: to move a pixel from 50 to 150, the scaler briefly drives it as if you had asked for 200, then drops back to 150 on the next frame. Done well, the pixel arrives at 150 in one frame instead of three. Done too hard, it arrives at 180, and that overshoot is visible as a bright halo trailing the object — inverse ghosting. The correct amount of push depends on the frame interval, which is why variable overdrive (a per-refresh-rate table, common on better VRR monitors) exists and why monitors without it are a compromise tuned for one refresh rate.
The blur floor no response time can beat
Even a hypothetical 0 ms panel blurs moving objects, because a normal display is sample and hold: each frame is painted once and then held perfectly still for the entire refresh interval while your eye keeps gliding along the object's path. The still image is therefore smeared across your retina. The size of that smear is simply speed ÷ refresh rate, which is the css px / frame figure in the readout: 960 px/s at 60 Hz is 16 px of blur, at 144 Hz it is 6.7 px, at 240 Hz it is 4 px. No monitor menu can touch that number. More Hz, or strobing, are the only two cures — which is exactly why a 240 Hz panel with mediocre pixels can still look clearer in motion than a 60 Hz panel with excellent ones.
Black frame insertion and backlight strobing
ULMB, ELMB, DyAc, MBR, "1 ms MPRT" — different names for the same trick. Instead of holding each frame for the whole interval, the backlight is pulsed for a short slice of it (or a black frame is inserted between real ones), so your eye only integrates a brief flash and persistence blur collapses. It genuinely works, and the cost is genuine too: you lose a large fraction of peak brightness, it usually cannot run at the same time as G-Sync or FreeSync, your frame rate has to stay pinned at the strobe rate or you get visible double images, and the flicker gives some people eye strain or headaches. Try it, but do not assume it is strictly better.
If the result is bad, what do you do about it
- Smear only on the dark rows — that is a VA panel behaving like a VA panel. Raising overdrive one step helps a little; nothing removes it completely, because the physics of a dark-to-dark transition are the trade you made for the contrast ratio.
- Halos everywhere — overdrive is too high. Drop a step. If the lowest non-off step already halos, use Off.
- Nothing changes as you step through overdrive — some monitors ignore the setting in certain picture modes, or lock it while VRR, HDR or a strobing mode is active. Reset the monitor to factory defaults and try again on a plain SDR picture mode.
- Clean here, bad in a game — the panel is fine and your frame rate is not. Ghosting caused by a monitor does not change when you turn shadow quality down; stutter does.
- Smear that got worse recently on a laptop — check for a panel power-saving or "panel self refresh" option in the GPU control panel, and make sure the machine is plugged in. Battery modes routinely drop the refresh rate and slow the panel.
FAQ
What is monitor ghosting?
Ghosting is a trail left behind a moving object because the pixel cannot finish changing colour before the next frame arrives. It is still on its way from the old shade to the new one when it is asked to change again, so a faded copy of the object smears along behind it. It is a property of the panel, not of your graphics card, and no driver setting removes it.
What is inverse ghosting or overshoot?
Overshoot is what overdrive looks like when it is turned up too far. To make a pixel change faster the monitor briefly drives it past the shade it is aiming for, then lets it settle back. If the push is too strong the pixel visibly overshoots and you get a bright or dark halo trailing the object with a hard edge, instead of a soft tail. Most people find overshoot more distracting than the smear it was meant to remove.
Which overdrive setting should I use?
There is no universal answer, because the right step depends on the panel and on the refresh rate you run it at. Run this test, start with overdrive off, then raise it one step at a time and stop at the last step before a bright halo appears behind the moving object. If your monitor lacks variable overdrive, check the setting again at 60 Hz, because a level that is perfect at 144 Hz often overshoots badly when the panel refreshes more slowly.
Is a 1 ms response time real?
Not in the way the box implies. The advertised figure is a best-case grey-to-grey number: one favourable transition, measured with overdrive at its most aggressive setting, using a 10 to 90 percent window that ignores the slow start and the slow finish. Averaged across the full range of transitions a real panel is usually several times slower. A 1 ms MPRT claim is a different number again, describing how long each frame stays visible with backlight strobing on rather than how fast the pixels move.
My screen stutters during the test. Is that ghosting?
No, and it is the most common way people mis-tune a monitor. A dropped frame makes the object jump, and the jump reads to your eye as a double image that looks a lot like a smear. Check the frame-rate readout under the strip: if it sits below your refresh rate, or the graph shows red bars, fix that before you touch the monitor menu. Close other tabs, quit recording and streaming software, use full screen, and plug a laptop in.
Do black frame insertion and backlight strobing get rid of blur?
They remove a different kind of blur. Even a perfect panel looks blurred because each frame is held on screen for the whole refresh interval while your eye keeps moving, which is sample-and-hold persistence rather than slow pixels. Strobing the backlight, or inserting black frames, shortens the time each image is visible and cuts that blur sharply. The price is real: you lose a large part of your brightness, it usually cannot run at the same time as variable refresh, and the flicker bothers some people.