FAQ
The questions creators actually ask
Every question below is one creators put into a search box, answered with the numbers this site can source. Nothing here is a sales pitch for a product that has not launched.
YouTube thumbnails
Sizes, file formats, the crop, and the parts of a thumbnail that decide whether anyone clicks it.
What size should a YouTube thumbnail be?
YouTube recommends 3840 x 2160 px at 16:9, with a minimum width of 640 px, saved as JPG or PNG. Between 1280 x 720 and 1920 x 1080 neither is wrong, because both are the same 16:9 frame and both clear the 640 px floor. 1280 x 720 is the largest thumbnail YouTube ever serves to a browser, the file known as maxresdefault.jpg, so everything above it is headroom for the encoder rather than pixels anyone sees. Export at 1920 x 1080 and you satisfy the recommendation, the minimum and the 2 MB mobile upload cap at once. The ratio is the only spec with no tolerance. Every number is on the thumbnail size record and in the size guide.
Why do YouTube thumbnails look blurry or low quality?
Blurry thumbnails are almost always compression, not resolution. YouTube re-encodes every upload and serves a fixed ladder of derivatives, 120 x 90, 320 x 180, 480 x 360, 640 x 480 and 1280 x 720, and the slot decides which one a viewer gets. Thin type, soft gradients and busy backgrounds are what fall apart first, so the fix is usually the design rather than a bigger file. Export at 1920 x 1080 or larger, put flat areas of colour behind any text, and set the words heavier. Before uploading, look at your own file at each of those sizes, including the 168 px sidebar, where the damage shows first.
Can I change a YouTube thumbnail after posting a video?
Yes, you can replace a YouTube thumbnail on a published video at any time, as many times as you like. Swapping it does not delete views, wipe watch time or restart the video's ranking, and YouTube's own help page says a re-rank after a change comes from how viewers respond rather than from the act of changing it, so there is no penalty for fixing a bad one. You do need a verified account before the custom thumbnail option exists at all. The real cost lands on a video that is already performing, because once the image changes you cannot compare like for like. If you expect to iterate, keep the layered source file, since a second version is only worth testing when you can edit the first. Put both candidates in a mock feed before you swap.
Should a YouTube thumbnail be JPG or PNG?
Use JPG for photographic thumbnails and PNG for flat graphics and hard-edged type. YouTube accepts those two formats and nothing else, so WebP, AVIF and SVG are rejected outright. The reason to prefer JPG for photos is weight. A 1920 x 1080 photographic PNG often lands at 3 to 6 MB, while the same image at JPG quality 80 sits under 1 MB with no visible difference at feed size. That matters because mobile uploads cap at 2 MB and YouTube's Data API caps at 2 MB as well, so any tool uploading on your behalf inherits the limit. Keep PNG while you work, especially for a transparent cutout, then flatten to JPG on export.
How long does it take for a new YouTube thumbnail to update?
YouTube publishes no time for a thumbnail to update, so treat every figure you have read as someone's observation rather than a commitment. The pattern people see is consistent: YouTube Studio and the watch page pick it up first, an already loaded feed keeps serving the old image until it refreshes, and search and suggested settle last. Re-uploading repeatedly does not speed any of that up, it just leaves you more versions to be confused by. Reload, give it a while, and check again. To see which file YouTube is handing out right now, pull the live image straight from the video, and if it still will not move, the four causes are here.
Do thumbnails matter on YouTube Shorts?
Yes, but only on the surfaces that show a still image. A Shorts cover appears in your channel Shorts tab, search results, hashtag pages, audio pages and the Shorts shelf in the home feed. In the vertical Shorts player the video starts immediately and the cover is never rendered, so your first second of footage does that job instead. They are two separate pieces of work. The cover is 2160 x 3840 px at 9:16 with a minimum height of 640 px, and exporting at 1080 x 1920 px is the sensible choice. Two catches: custom Shorts covers can only be added in YouTube Studio on a computer, and since they shipped on 24 July 2026 they have been limited to YouTube Partner Program creators with a verified account while YouTube expands access. Details are on the Shorts cover size record.
Can I make YouTube thumbnails on my phone?
Yes, and for a long-form video you can do the whole job on a phone. Two limits are worth knowing. A thumbnail uploaded from the YouTube mobile app is capped at 2 MB against 50 MB from desktop, so a heavy PNG that gets refused on your phone is the cap, not a bug. And a custom Shorts cover can only be uploaded in YouTube Studio on a computer, by a YouTube Partner Program creator with a verified account, so that one genuinely needs a laptop. The free tools on this site run inside the browser with nothing uploaded, so they work on a phone, and the resizer gets a file under the cap. DesignerOP's builders are prelaunch, waitlist only.
How much text should a YouTube thumbnail have, and where should it go?
Three or four words, kept in the middle of the frame and never in the bottom right corner. The title underneath already carries the sentence, so repeating it prints the same words twice on the watch page. Scale is why the count is so low. A headline set 100 px tall on a 1280 px canvas renders around 25 px in a feed and about 9 px in the smallest slot YouTube serves. Stay clear of the duration stamp bottom right, the red watched bar along the bottom edge and the hover buttons top right, with roughly a 6 percent margin all round. The safe zone checker draws every overlay at real size.
What makes a good YouTube thumbnail, and what mistakes should I avoid?
A good thumbnail has one subject, one promise, and a focal order you can still read at sidebar size. The mistakes are consistent: setting the video title as the thumbnail text, running the headline into the bottom right where the duration stamp lands, reusing a 16:9 image on a Short, and never once looking at the file small. The 2026 pattern is subtraction, fewer words and calmer frames, because contrast is relative and a quiet frame stands out in a row of loud ones. Test the headline with a real contrast and legibility check, and see what changed this year in the 2026 trends piece.
What do YouTubers actually use to make their thumbnails?
Most channels shoot their own photo, cut the subject out, and set two or three huge words over it in a layered editor, then let YouTube's built-in A/B test pick the winner. The specific tool matters less than people assume, because the expensive part is not making the first version, it is fixing one word in it. That test, called Test & Compare until YouTube renamed it, runs up to three titles or thumbnails on the same video and decides by watch share, and it is free, which turned the job from making one thumbnail into making variants that change one variable each. We compared the options in the makers roundup, including ones we would pick over our own builder, which is still prelaunch.
AI, tools and what this costs
What a generator can and cannot do, why AI text comes out wrong, and where DesignerOP sits among the tools you already know.
Can AI make YouTube thumbnails?
Yes, and the useful question is not whether AI can make one but what it hands back. A prompt generator returns a single flat image where the subject, the background and the words are welded together, so changing one word means generating the whole picture again. Reference based generators start from a thumbnail you already like, which is a better starting point, though several still return a flat file where an edit is billed like a fresh generation. A layer based tool returns background, subject and text as separate objects you can move on their own. We sorted the field by that one question, and whatever you use, check the result at the sizes it really renders before you upload.
Can ChatGPT make a YouTube thumbnail?
ChatGPT can generate a 16:9 image you could upload as a thumbnail, but it returns a flat picture rather than an editable design. Any words in that picture are painted pixels, so spelling is a gamble and fixing one letter means regenerating the frame, which moves the subject and the lighting with it. It also has no notion of YouTube safe zones or whether your type survives at sidebar size. The workable version is to ask it for the picture only, never the headline, then set your words as real type in any editor with a text tool. Here is why the letters fail, and here is your result in a mock feed.
Why does the text in AI-generated thumbnails come out garbled or misspelled?
Because diffusion models draw letters as texture rather than language. The encoder that reads your prompt works in tokens, not characters, so the model never sees an S as an S, and it is scored on the whole image, where your headline is a thin sliver of pixels. Skin, sky and lighting win that trade every time, which is why AI text is rarely random noise and usually nearly right, with one letter doubled or wrong. Prompt tricks like quotation marks, shorter words or asking for a sign change the odds, not the mechanism. The long version is here. The real fix is to keep words out of the image model and set them as real type, which you can preview over your own image.
Is there a genuinely free thumbnail maker with no watermark and no sign-up?
Yes for checking and preparing a thumbnail, and rarely for AI generation, because generations cost real money to run. Every tool in the DesignerOP free tools section runs entirely in your browser, needs no account, adds no watermark and never uploads your image: feed previews, an A/B mock test, safe zone and contrast checks, font pairings, palette extraction, background removal and a carousel splitter. For actually building one at no cost today, the template paths in Canva and Adobe Express have free tiers and real editable text. DesignerOP's own builders are prelaunch and not open to the public, and there is no free tier planned.
Do I need Photoshop or design skill to make a thumbnail?
No, you can make a thumbnail that works without Photoshop and without design training. Photoshop passes every edit test ever devised, but it costs a subscription and the skill to use it, and it turns a ninety second job into an evening while you learn. What decides whether a thumbnail works is four decisions, not software: one expressive subject holding an emotion a stranger can name, one simplified high contrast background, three to five words with a single punch word, and one obvious focal order. Those are choices you can make in any editor that has layers and a text box. Rebuild a reference you already like instead of inventing a composition, and pull its colours out as hex codes rather than eyeballing them.
Which is the best YouTube thumbnail maker?
No thumbnail maker is best for everyone, so choose the one by what fixing a single word costs you. Publishing this week with no budget: Canva templates or Adobe Express, both real layers with real type and a free tier. Want AI from a reference today: Pikzels or vidIQ, with eyes open, since vidIQ charges the same 22 credits for a generation, a regeneration or an edit. Already deep in YouTube tooling: TubeBuddy, a real layer editor on a cheap entry plan. Full manual control: Photoshop. The full comparison is here. DesignerOP is prelaunch, so it is not an option today. Before you publish, put two candidates in the same feed row.
How is DesignerOP different from Canva?
Canva offers three different routes to a thumbnail and DesignerOP offers one. Canva templates give you real layers and real type from a library thousands of channels share. Magic Media takes a prompt and returns a flat raster with the words painted in. Magic Layers, announced in March 2026 as a beta in the US, UK, Canada and Australia, converts a flat image back into layers, metered against your AI allowance, and beta testers report the split is best effort on stylised type and soft edges. DesignerOP starts from a reference and builds structural layers whose text was never pixels in the first place. Four cases where Canva wins, and the wider map of the field.
What is a reference-driven thumbnail maker, and is copying someone's thumbnail allowed?
A reference driven maker starts from a thumbnail that already works and rebuilds its structure, instead of asking you for a prompt or a blank canvas. The structure is what you take: where the subject sits, how contrast separates it, how many words there are and which one is biggest. That is ordinary practice, and nobody owns big face right, three words left, flat background. The skin is what you leave: the photograph, the artwork, the person and their face. Someone owns those, and they advertise someone else's video. If your finished thumbnail still contains a single pixel from the reference, you did not rebuild it, you reposted it. The step by step rebuild is here.
What does DesignerOP cost, and when does it launch?
Creator is $19 a month with 100 credits and Studio is $39 a month with 200 credits, and there is no launch date to give you. Annual billing takes 30% off both, which works out to $13.30 a month billed $159.60 a year for Creator, and $27.30 a month billed $327.60 a year for Studio. One credit is one generation, editing is free forever, unused credits roll over capped at 200 on Creator and 400 on Studio, and a failed generation refunds its credit automatically. Top ups are $9 for 50 credits, for subscribers. There is no free tier. Pricing is published in full while the product is still prelaunch with an open waitlist, and payments open at launch.
Is the text in DesignerOP really editable, or is it pixels in an image?
The text in DesignerOP is real, editable type, never pixels in an image. Words you type are genuine text objects in a real font, and they never enter the image model at all, which is why the output cannot come back misspelled the way generated letterforms do. On the design surfaces the background, the subject and the text are independent layers, so retyping a headline re-renders nothing and costs no credit, and nudging the subject leaves the other two where they are. Being straight about it: DesignerOP is prelaunch, so this describes how the thumbnail builder is built and what we hold ourselves to, not something a stranger has tested. The architecture is written out here.
Instagram carousels, Reel and TikTok covers
Slide counts, the ratio lock, the grid crop, and turning one video into a deck.
How many slides should an Instagram carousel have?
Six to ten slides is the working range for most carousels, and the hard limits are 2 to 20 in the Instagram app. Meta's publishing API caps a carousel container at 10 items, so a set of 15 built for a scheduling tool will be refused past the tenth slide. Six to ten also matches how the format is read: one hook slide, one claim per slide after it, one slide that asks for something at the end. Only slide one becomes your profile grid tile, so the extra slides earn swipes, not grid space. The full carousel spec has the slide count sources.
What size should an Instagram carousel be, and why does Instagram crop my photos?
Build every slide at 1080 x 1350 px, a 4:5 portrait. The crop you see is the ratio lock rather than a height limit: Meta's publishing documentation says carousel images are all cropped based on the first image, so one square slide one squashes every slide behind it. The feed itself is more generous than most guides say. Adam Mosseri announced on 28 May 2025 that Instagram supports 3:4 photos, 1080 x 1440 px at that width, and 4:5 and 1:1 were kept rather than replaced. The second crop is your profile, where the 3:4 grid tile narrows a 4:5 post to 1012 px wide. Keep type inside a 1000 px column and both crops leave it alone. Every number is on the portrait post record.
How do you make a seamless Instagram carousel where the image lines up across slides?
Cut one wide image into equal vertical strips, export every strip at the same ratio and 1080 px wide, then post them in slide order. The ratio has to be identical on every slice, because Instagram crops the whole set to match slide one, and that single mismatch is what breaks most hand made attempts. DesignerOP's free carousel splitter does the cutting in your browser with no account, at 2 to 10 slides. Be realistic about the effect though: Instagram shows one full slide at a time with no peek at the next, so every seam is a hard cut. Three to five slices read best, and a repeated motif holds a set together more reliably than a split background.
Do Instagram carousels get shown to people who do not follow you?
Yes, Instagram shows carousels to people who do not follow you. Carousels surface beyond your followers through the Explore tab and through hashtag and topic result grids, not just in the feeds of people who already follow you. What a stranger sees there is only slide one, cropped to the 3:4 grid tile, so slide one is the entire pitch. Reach is not the carousel's strong suit though. Buffer's 2026 analysis of 52 million posts found Reels are the best format for reach, while carousels earn a median engagement rate of 6.90% of people reached against 3.31% for Reels. More people see a Reel, a larger share of them act on a carousel. The numbers and sources are in carousels vs Reels.
How do you turn a video into an Instagram carousel?
Pull the transcript, cut it down to six to ten claims, write the hook from the sharpest one, put one claim on each slide, then design one layout and repeat it across the set. It is a rewriting job, not a cropping job. Screenshotting frames fails on four counts: a 4:5 slide keeps only about 45% of a 16:9 frame's width, a 1080p frame cropped to 4:5 yields at most 864 x 1080 so every slide is an upscale, burned in captions turn to mush, and a frame from minute four means nothing to a new reader. DesignerOP's video-to-carousel maker drafts that set at 1080 x 1350 and hands it to a layer editor. The step by step is here, and the maker is prelaunch with an open waitlist.
What size is an Instagram Reel cover, and how do I stop the grid from cropping it?
An Instagram Reel cover is 1080 x 1920 px, 9:16, which is the ratio Meta recommends for Reel cover photos as a JPEG up to 8 MB in sRGB. You cannot stop the crop, but you can decide what falls inside it. The main feed shows a 4:5 slice at 1080 x 1350 px, taking roughly 285 px off each end, and the profile grid shows a 3:4 tile at 1080 x 1440 px, taking about 240 px. Keep everything readable inside the central 1080 x 1350 px band and both crops keep it. The cover editor also has a separate profile grid step where you drag the tile on its own. Check your own file against every crop with the cover size previewer, or read the full cover spec.
What size is a TikTok cover photo, and why does mine look blurry or refuse to update?
A TikTok cover is 1080 x 1920 px, matching the 9:16 video frame. Blur almost always comes from scrubbing to a frame instead of uploading a still. That frame was compressed as part of the clip, so any text burned into it is softer than the same words on a separate file, and it carries whatever motion blur the moment had. The cover is set on the post screen, so it is worth deciding before you post. TikTok does not document whether it can be changed afterwards. Instagram is the clearer case: a Reel cover can be edited once the Reel is live, from the three dot menu on the post. Sizes and safe zones are on the TikTok cover page.
Do Reel and TikTok covers matter, or does the algorithm ignore them?
Covers matter for the tap, and nobody outside Meta and TikTok knows what ranking weight, if any, they carry. That is the honest split, and the tap is the half you control. Your cover is what a stranger meets in every grid surface: your profile, Explore, search results, sound and hashtag pages, where your post sits directly beside other people's. In the player it is only on screen for a moment, but on those grids it is the whole post. On Instagram the cover stays editable after the Reel is live, from the three dot menu on the post, which makes it the cheapest thing on a post to change. What that change does to reach is not something Meta publishes, so nobody can promise it is free. Why vertical covers are built as layers is explained on the cover builder page.
Can one design be resized for Instagram, TikTok and LinkedIn without redoing it?
One 1080 x 1350 px design covers Instagram and LinkedIn unchanged, but TikTok needs a separate vertical file. 4:5 is the tallest ratio LinkedIn allows, since LinkedIn takes 3:1 through 4:5 and recommends a 1080 px width. Instagram posts 4:5 uncropped too, and since 28 May 2025 it also takes a taller 3:4 photo, so 4:5 is the one shape that fits both. Watch the file caps: 8 MB on Instagram, 5 MB per photo on LinkedIn. TikTok and Reels play at 1080 x 1920, which is a different composition rather than a resize, so the subject and the words have to sit in the middle band. DesignerOP's free image resizer crops one file to eight sizes around a focal point you set, in the browser. LinkedIn's own limits are on the post image page.
What is a carousel post, and when should a creator use one instead of a Reel?
A carousel post is one Instagram post holding several images that people swipe through, up to 20 in the app and 10 when posted through Meta's API. Use one when the idea survives as six to ten one line claims, which usually means a lesson rather than a moment. The two formats do different jobs. Buffer's 2026 analysis of 52 million posts found carousels earn a median engagement rate of 6.90% of people reached against 3.31% for Reels, while the same research finds Reels are the best format for reach. Metricool's 2026 study found carousels get nine times more saves than single image posts. Reels find strangers, carousels convert them, and the full comparison covers running both.
Testing, CTR and things going wrong
What the numbers in Studio actually mean, how YouTube's own thumbnail test decides, and what to do when a thumbnail will not show up.
Why is my YouTube thumbnail not showing, or not updating after I changed it?
A YouTube thumbnail that will not show or update has four separate causes, and telling them apart saves hours of re-uploading. No upload option under Thumbnail means the account is not phone verified, the gate YouTube puts on custom thumbnails. An upload that bounces back is usually the format, or the cap for the device you uploaded from, 2 MB on mobile against 50 MB on desktop. A daily custom thumbnail limit message is a per channel cap with no published number, only advice to try again in 24 hours. A thumbnail that vanished alongside an email is a Community Guidelines removal, so appeal before replacing it. The triage order is here, and you can pull the live file off the video to see what YouTube is holding.
Why does the old thumbnail still show when I share the link on WhatsApp, Discord or Facebook?
The old thumbnail in a WhatsApp, Discord or Facebook preview is that app's cached copy, not something YouTube is still serving. A watch page carries an og:image tag pointing at an address built from the video ID, so changing the thumbnail changes the picture at that address, never the address itself. Those apps key their preview cache on the address, so a link they have seen before is never fetched again. Meta gives you one lever, its Sharing Debugger, where pasting the watch URL and pressing Scrape Again corrects Facebook and Instagram. Elsewhere, hand the app a link it has not seen. Previews already sitting in other people's chat histories cannot be flushed at all, so settle the image before the first share. The mechanism is written out here.
What is a good YouTube thumbnail click-through rate?
YouTube publishes one benchmark for thumbnail click-through rate: half of all channels and videos have an impressions CTR that can range between 2% and 10%. That band holds half the platform, so it is not a target and not an average. A quarter of videos sit below it and a quarter above, and plenty of both are healthy. YouTube attaches two caveats, that channels or videos less than a week old and videos under 100 views can land well outside it, and that a video collecting a lot of impressions, for example on the home page, will naturally show a lower rate. The tidy 4% to 5% average quoted everywhere is not YouTube's figure. The full reading is here, and you can check the thumbnail at the size it renders.
How do I check the click-through rate of one specific thumbnail?
The click-through rate of one specific thumbnail lives on the Reach tab of that video's analytics, not on the channel overview. In YouTube Studio on a computer, open Content, select Analytics on the video you want, then choose Reach. Two numbers there matter: thumbnail impressions, meaning how many times your thumbnails were shown to viewers, and impressions click-through rate, meaning how often viewers watched after seeing one. Know what the denominator leaves out before you trust it. YouTube counts an impression only if the thumbnail was on screen for more than one second with at least half of it visible, and it counts nothing from external sites, mobile web, YouTube Kids or background tabs. Judge the image itself in a mock feed and at sidebar size.
How long should I wait before changing a thumbnail?
The wait before changing a thumbnail is measured in impressions, not days. A few hundred impressions tells you nothing, a few thousand starts to say something, and below roughly ten thousand small differences are noise. That threshold comes from arithmetic rather than from YouTube: a true 5% rate can print as anything from 2.5% to 7.5% at 300 impressions, and only tightens to 4.6% through 5.4% at 10,000. YouTube's own guidance agrees in plainer words, listing deciding without enough data as a mistake and warning against acting on small variations. A swap also spends the comparison, since views before and after pool into one line. The impressions maths is here, and a side by side mock feed helps you choose before you publish.
How does YouTube's Test and Compare decide which thumbnail wins?
YouTube's thumbnail test, now called A/B test titles and thumbnails and formerly Test and Compare, picks a winner by watch time share rather than click-through rate. Its help page says the result is based on watch time share, and that tests are optimised for overall watch time over other metrics like click-through rate, because a thumbnail should also tell a viewer what the video is. There are three outcomes: Winner, Performed Same, and Inconclusive, where the first thumbnail you uploaded becomes the default. Up to three options run concurrently, with a small percentage of traffic held back as a control group excluded from the maths. That is why a paid tool ranking by clicks can crown a different winner. Variants differing on one real idea resolve fastest, which is easier when the file stays layered. The mechanics in full.
Why can't I see A/B testing on my channel?
A/B testing hides behind four separate gates, and missing any one of them removes the feature. It is only available on computers through YouTube Studio, so the mobile Studio app will never show it. It sits under advanced features, which need either sufficient channel history or ID or video verification. Some formats are excluded outright, since Shorts, Scheduled Lives and Premieres cannot be tested, though Live Archives can and a Premiere qualifies once it ends. And it is unavailable on videos made for kids, made for mature audiences, or set to private. Check Settings, then Channel, then Feature eligibility in Studio on a computer. Every gate is written out here, and a browser mock feed compares candidates while you wait.
Can you A/B test Shorts covers, Instagram Reel covers or TikTok covers?
YouTube Shorts, Instagram Reels and TikTok all lack a native cover test, so the comparison is yours to make before publishing. YouTube says it outright: A/B testing is not available for Shorts, and once a video transitions to a Short you lose any test created for it. You can still set a Shorts cover, just not test one, and only in Studio on a computer. Instagram's closest feature is Trial Reels, which shares a whole reel with non-followers for early feedback, not two covers against each other, and TikTok documents nothing comparable. Our A/B tester is a browser mock feed for picking between candidates, not a connection to any platform's test. Check each candidate against the real crops, and the Shorts cover spec has the sizes.
Specs on this page were checked against platform documentation on 2026-09-04. A number that goes stale gets corrected, so mail hello@designerop.com if you find one.