MrBeast Thumbnails: What Is Actually Documented, What Is Folklore, and What Transfers to a Channel Your Size

Search for MrBeast thumbnails and you will be told, with total confidence and no source, the hex values of his palette, the name of the typeface his team uses, the exact dollar figure he spends per image, and the percentage lift you can expect from copying any of it. Almost none of that is sourced. Some of it traces back to a comedy sketch. The rest is one blog quoting another blog quoting a tweet.

The documented record is much smaller than the folklore, and much more useful, because what is actually documented is a process rather than a style. A process is the only part of a thumbnail department with storyboards and stand-ins that a one-person channel can run at home. The style is the part that has been copied so widely that it now works against the channels copying it.

This post separates the two. What his own team has said, what the leaked production guide says, which of the famous numbers survive a check, which parts of the approach are free and transferable, and why the ending of this argument was written in June 2025 by MrBeast's own company.

What his own team has actually said

The strongest source on how MrBeast thumbnails get made is not a creator blog. It is an interview YouTube itself published with Chucky Appleby, the designer who leads thumbnails for the channel. Social Media Today's write-up of that interview reports a team that arrives at roughly fifty thumbnail and title concepts for a single video before whittling them down to the combination they ship.

Two things in that account matter more than the number. The first is that thumbnails and titles are generated together, as pairs, not as an image that gets a caption later. The second is the criterion used to cut them down. In coverage of the same interview, Appleby's stated test is the person scrolling: the frame has to let someone choose the video, and complexity is what loses, because a thumbnail that takes work to decode gets scrolled past rather than decoded.

That is a far less glamorous answer than the style guides imply. It is also the answer that agrees with the measurement work: at the size a thumbnail is actually judged — a few hundred pixels wide, for a fraction of a second, inside a grid — the number of things the eye has to resolve is the variable that decides whether the frame communicates at all. The composition post works through what survives that shrink and what does not.

The leaked production guide, and the one line in it that matters

In September 2024 a 36-page internal document titled How to succeed in MrBeast production circulated widely. Coverage at the time, including Sherwood News, treated it as genuine but noted it was never confirmed as authentic by the company, and accounts of how it surfaced differ. Read it as a reported document rather than a published one.

On packaging it says something simple and more important than any design tip in it: the title and thumbnail on the video you are producing set the viewer's expectations for that video. The consequence, which the guide draws out, is that packaging is a production input. If the frame promises a thing, the shoot has to deliver that thing, which means the frame has to exist before the shoot.

The guide also names the three numbers staff are told to watch: click-through rate, average view duration, and average view percentage. That triad is worth more than it looks. It rules out the failure mode that every thumbnail guide on the internet encourages — optimising the click in isolation — because a frame that wins clicks and loses the first minute shows up immediately as CTR rising while AVD falls. Our post on what counts as a good CTR covers why the ratio is meaningless read on its own.

The other packaging note reported from the document is a push towards the more extreme phrasing of the same truth: the example given is titling a video I Survived rather than I Spent. This is the line creators quote as licence to exaggerate, and it is worth being precise about what it licenses. Both titles describe the same video. One describes the stakes and one describes the activity. That is a selection between true framings, which is a different act from promising something the video does not contain — the distinction the clickbait post spends its length on, because the second version is where retention and policy problems both start.

A thumbnail studio is a repeatability machine

In 2024 Jon Youshaei published a tour of what he described as MrBeast's thumbnail studio, and the details he listed are the most instructive thing in the whole corpus: storyboards used to pick concepts, stand-ins used to fix positions before the talent arrives, and gridlines on the floor so lighting and shadow can be reproduced on a later shoot.

None of that is about making one beautiful image. All of it is about making the twentieth image as consistent as the first, at speed, with different people doing the work. The channel's visual recognisability — which is the actual asset — is manufactured by removing variance from the process.

The solo-creator equivalent is unglamorous and nearly free. It is a layered file where the background treatment, the type block and the subject slot are already placed, so each upload is a swap rather than a design. It is a saved camera position and a light you do not move. It is the same crop, the same type size, the same two accent colours for a year. The consistency post covers how much of that is worth locking and which parts should stay free.

Which famous numbers have a source, and which do not

Four claims make up most of what is written about MrBeast thumbnails. They are not equally supported, and the difference matters if you are about to spend money because of one of them.

Claim	Where it traces to	Status
Around fifty thumbnail and title concepts per video	Coverage of YouTube's interview with Chucky Appleby, the channel's thumbnail lead	Best supported claim in the set, though via secondary write-ups of the interview rather than a transcript
$10,000 spent per thumbnail	Secondary articles and a retelling of a comedy video; no company disclosure	Unverified. Widely repeated, never sourced
Twenty thumbnail variations per upload	Background in later write-ups, sitting alongside the fifty-concepts figure it contradicts	Unverified, and inconsistent with the other number
A specific named typeface is “the MrBeast font”	Design blogs naming one of two or three heavy display faces, each asserting it confidently	Unverified. Treat font lists as sampled guesses

The font claim is the one that can cost you directly, and not because of the aesthetics. Free display faces are not uniformly licensed for commercial use, and a monetised thumbnail is commercial use; the fonts post covers which licences actually permit it and what the legible minimum size works out to at feed width. Copying a font you were told belongs to a channel is a worse reason to pick type than any of the reasons in that post.

The spend figures deserve one more sentence, because they are quoted at small channels as a benchmark. What MrBeast's thumbnail budget tells you about your own is nothing. A frame in front of tens of millions of impressions justifies a cost per image that would be irrational at a thousand. The designer cost post does that arithmetic from the other end: what an image is worth to you, given your impressions.

Thumbnail first is the transferable part

Strip the budget, the studio and the staff out, and what is left of the documented approach is an ordering rule: decide the title and the frame before you film. It is the cheapest change in this entire post and the only one that reliably moves outcomes, because it operates one stage earlier than packaging normally does.

Deciding packaging first does two things. It kills videos early: if you cannot draw the frame or write the title, the idea has no promise in it, and finding that out before a shoot costs a notebook page instead of a weekend. And it changes the shoot: once the frame exists on paper, you know you need a wide of the room before it is wrecked, the reaction shot at the moment of the reveal, and the object held at chest height against a clean background. Creators who package last are usually searching exported footage for a usable frame that was never filmed.

The corollary is a shot list item rather than an edit decision. Film the thumbnail. One or two minutes of deliberate capture — subject lit, background clean, framing loose enough to crop to 16:9 — is worth more than any amount of retouching applied to a screen grab. Our post on shooting photos for thumbnails covers the lighting and lens side of that.

Fifty concepts is not fifty designs

The fifty-concepts figure gets misread as fifty finished images, which is both impossible for a solo creator and the wrong lesson. A concept is a pairing: a promise and the image that carries it. Most can be a sentence. Only a handful ever need rendering.

The version of that process which fits one person and an afternoon:

1. Write ten packaging lines for the video, each a title plus a one-sentence description of the frame. Prose, not design.
2. Cut every line whose frame could belong to any of the other nine. Convergence is the enemy, and it shows up in the sentence before it shows up in the image.
3. Keep the three that promise different things — not three executions of one promise. One stakes, one outcome, one question.
4. Build two of the three properly, at full size, with the type set.
5. Check both at feed size in the free thumbnail preview tool before you form an opinion at 100% zoom, where nobody will ever see them.
6. Put them head to head in the A/B comparison tool, and if the video is eligible, ship the pair into YouTube's own test instead of guessing.

If the bottleneck is step four — and for most channels it is, because building two real candidates is an hour of work that competes with editing — that is the specific problem a generator is for. Our prompting post makes the same argument from the tooling side: brief for three, pick one, never re-roll the same prompt hoping for luck.

What MrBeast thumbnails actually do at feed size

Set the folklore aside and look at the frames as compositions. The recurring structure is not mysterious: one subject that reads as a silhouette, a second element that supplies scale or stakes, two or three words at most, a palette narrow enough to be identifiable in a column of nine, and a background with nothing competing in it.

That list is ordinary good practice. It is what the colour post and the composition post arrive at from the measurement side, and it is what a well-built frame in any niche does. The honest summary of the visual half of this question is that there is no secret in it. His frames are competent executions of well-understood rules, produced at a volume and consistency almost nobody else can match, in front of an audience that already recognises them.

What is specific to him — the open-mouthed expression, the extreme saturation, the money-and-props tableau — is style, and style is the part you should be most reluctant to borrow.

Why copying the MrBeast thumbnail look is now the weakest move available

Distinctiveness is relative, not absolute. A thumbnail is never judged on its own; it is judged against the eight to eleven frames beside it, and the whole value of a loud style is that the neighbours are quieter. When the neighbours adopt it, the signal inverts: the saturated close-up with the oversized number becomes the background, and the quiet frame becomes the one that stops the scroll.

This is no longer a prediction. The style's ubiquity is now the thing critics name when they write about YouTube's homepage — Creative Bloq, covering the AI-tool episode described in the next section, wrote of shocked faces, dodgy Photoshop and oversaturated colours having plagued the homepage for years. Whatever you make of that as criticism, as a description of a competitive field it is actionable: the look is saturated, and saturated looks do not differentiate.

There is a second, quieter reason. His frames are optimised for a job yours probably is not. At his scale the overwhelming majority of impressions come from browse surfaces served to people who have no relationship with the channel, and the frame's job is to stop a stranger. A channel with a returning audience is solving a different problem: being recognised fast, and being trusted. Those pull in opposite directions — recognition wants sameness, stopping a stranger wants novelty — and which one you need is answered by your own traffic mix, not by his. The traffic sources post covers how to read that split in Studio.

The ViewStats episode, because it is the end of this argument

On 20 June 2025, ViewStats — the analytics company MrBeast co-founded — launched an AI thumbnail feature. As reported at the time, it offered face swapping and the ability to use any YouTube channel as a style reference, on a subscription of roughly $80 a month.

The reaction from creators was immediate and specific. The objection was not that AI had been used; it was that the product's selling point was reproducing other people's thumbnail work. The streamer PointCrow's complaint, quoted by the BBC in its coverage, was that the model was clearly trained on creators' thumbnails and used them without permission. Others circulated a 2021 post in which MrBeast himself had objected to another channel copying his thumbnails.

Within days the feature was pulled. In a video addressing it he said he had “definitely missed the mark”, said the intention had been to help smaller creators, and noted he had spent more on thumbnail artists than any other YouTuber. ViewStats replaced the feature with a route to hiring a human designer.

The reason that episode belongs in a post about his thumbnails is that it is the style-copying question answered by the person whose style everyone copies. A tool whose premise was make my thumbnail look like that channel's was rejected inside a week by the community it was aimed at, and withdrawn by its own owner. If the most resourced thumbnail operation on the platform cannot make style transfer acceptable, a small channel building its identity out of borrowed style is not making a technical mistake. It is making a positioning one.

We build an AI thumbnail generator, so the distinction matters to us commercially and we will state it plainly: the useful thing a generator does is produce your frame faster — your face, your subject, your palette, three variants instead of one, on an afternoon when you would otherwise have shipped the default frame. The thing it should not do is hand you a reconstruction of a specific channel's artwork. The first is production capacity. The second is the product that got pulled.

Testing is how you find out, and the tooling got better

Every claim in every MrBeast thumbnail guide, including this one, is a hypothesis about your audience that your audience can settle. YouTube's own experiment tool compares up to three thumbnails on real traffic and decides on share of watch time rather than clicks, which is the right target: it is immune to the frame that wins the click and loses the viewer. The mechanics, the eligibility rules and the traps are in our thumbnail A/B testing guide.

Two things about the state of this are worth knowing. At Made on YouTube 2026 the company said creators have run more than 40 million title and thumbnail experiments since Test & Compare launched in 2024, which makes it the single most-used packaging tool on the platform and an odd one to still be arguing about. And the same announcements extended the idea past packaging: a feature for testing up to three different cuts of the same video, scored the same way, which coverage reports as arriving broadly in 2027. The hook will become as testable as the frame.

One discipline carries across all of it. Change one variable per test. A candidate with a new expression, a new background and a new title tells you only that something differed, which is the one result you cannot act on.

What to take and what to leave

Take the ordering: title and frame before the shoot, as a production input rather than a finishing step. Take the pairing: never design a frame without the title it has to work beside. Take the volume, scaled honestly — ten sentences, three concepts, two builds, not fifty renders. Take the repeatability, which is the real content of a thumbnail studio: a template, a saved setup, a palette you keep. Take the three-metric habit, so a click you won can be checked against the minute you kept.

Leave the look. Leave the hex codes someone sampled off a screenshot, the font named with no source, and the $10,000 figure that has been quoted into existence. Leave, above all, the assumption that the packaging of the largest channel on the platform describes a problem you have. His frame has to stop a stranger at a scale you do not operate at; yours probably has to be recognised by someone who already likes you, or to be the one quiet, legible thing in a row of shouting.

If the constraint is that building two real candidates per upload costs an hour you do not have, that is a production problem rather than a taste problem, and it is the one Thumblore exists to remove: your subject, your palette, several frames to choose between, in the time it used to take to make one. What to promise, and whether the video delivers it, stays yours. Those were always the decisions that mattered — which is the one thing every MrBeast thumbnail guide, including the unsourced ones, gets right.