Every tool makes thumbnails. This one grades them.
Paste a YouTube link and the thumbnail gets pulled and measured: does it survive being shrunk to feed size, is the contrast strong enough, does the duration badge sit on top of something that mattered. Measured facts about the image, not a made-up click-through score.
or
Real measurementsNo invented scoresNo sign-up
Fetching and measuring…
The thumbnail
How it looks where people actually see it
360px · desktop grid
246px · mobile feed
168px · sidebar
If you cannot read it in the smallest one, most of your audience cannot either.
What this measures, and what it deliberately does not
Plenty of tools will give a thumbnail a score out of 100 and a predicted click-through rate. Those numbers are invented. Click-through depends on your title, your audience, the topic, what else is on the page and when it was posted. No tool looking at a single image can know any of that, and one that claims otherwise is selling confidence rather than information.
So this page reports only what can actually be measured from the pixels, and names the threshold it is comparing against in every case. What you do with that is your call.
The checks
Check
How it is measured
Dimensions and aspect
Compared against YouTube's 1280 by 720 recommendation and 16:9. Anything else gets letterboxed or cropped somewhere.
Feed-size detail
The image is rendered down to 168 by 94, the sidebar size, and the mean absolute Laplacian is computed on it. That is a standard edge-energy measure: it falls when shapes and text turn into mush at small scale.
Global contrast
Standard deviation of luminance across the whole image, on a 0 to 100 scale. Low values mean a flat image that recedes in a busy feed.
Duration badge collision
YouTube stamps a timestamp over the bottom-right corner. That region is checked for edge density; a high value means something detailed is about to be covered up.
Edge margins
The outer 4 percent is checked for detail, because different surfaces crop thumbnails slightly differently and content parked at the very edge is the first thing to be lost.
Tonal range
The share of pixels that are nearly black or nearly white. Large amounts of either mean detail is already clipped and will not survive further compression.
Why feed size is the only size that matters
A thumbnail is designed at 1280 by 720 and almost never seen at 1280 by 720. In a desktop sidebar it renders around 168 pixels wide. On a phone it is a little larger but sits further from the eye. The practical effect is the same: roughly a seventh of the linear detail you were looking at while designing it.
This is why the usual advice about thumbnails is what it is. Three or four words maximum, because more will not be read. Heavy type, because thin strokes vanish. One clear subject, because a busy composition becomes texture. None of that is style preference; it is a consequence of the display size.
The three renders above are the actual sizes. Look at the 168 pixel one and ask whether you can tell what the video is about. That single question is worth more than any score.
What to do about a poor result
Low feed-size detail. Fewer elements, bigger. Cut words, increase the type size, and put a solid shape behind text rather than relying on the photo underneath.
Low contrast. Separate the subject from the background: cut it out and place it on a contrasting colour, or darken the background behind light text. The palette extractor reports which of your colours keep text readable.
Duration badge collision. Move whatever is in the bottom-right corner. The badge is about 12 percent of the width and always there.
Content in the outer margin. Pull it inwards. The safe zone guide shows where the boundaries actually fall.
Autopsy Questions
Does this predict my click-through rate?
No, and it will not pretend to. Click-through depends on your title, audience, topic and timing, none of which are visible in the image. This page measures properties of the picture itself and tells you the threshold each one is being compared against, so you can judge for yourself.
Which thumbnail does it pull from a YouTube link?
The highest resolution YouTube has published for that video, normally maxresdefault at 1280 by 720. If the uploader never provided one, it falls back to hqdefault, and the report says which one it used, because the lower-resolution fallback is letterboxed and will affect the aspect check.
What is the Laplacian measure in the feed-size check?
A standard edge-detection filter. Applied to the shrunken image and averaged, it gives a single number for how much definition survives at that size. Crisp text and clean shapes keep it high; fine detail that blurs into a smudge sends it low. It is a measurement of the image, not an opinion about the design.
Can I analyse a thumbnail before I publish it?
Yes. Use the "analyse an image file instead" option and pick the file from your device. It runs exactly the same checks, and that file is never uploaded anywhere: it is read straight into a canvas in your browser.
Is the YouTube thumbnail I paste sent to your server?
The image is fetched through a same-origin proxy on our side, because browsers will not let a page read the pixels of an image loaded straight from another domain. Only the public thumbnail URL is fetched, nothing is stored, and files you pick from your own device skip the proxy entirely.
Why does it flag the bottom-right corner specifically?
Because YouTube draws the video duration there on every thumbnail, in every surface, and you cannot turn it off. Anything you place in that corner is partly hidden. The check measures how much detail is in that rectangle so you know whether it costs you anything.
Does a good result guarantee a good thumbnail?
No. It means the image is technically sound: legible when small, well separated in tone, nothing important hidden. Whether the idea is compelling is a judgement no measurement makes for you. Passing every check and still being boring is entirely possible.
Continue crafting
Keep going without re-uploading — your image carries over to the next tool automatically.