How to make a thumbnail is a question about decisions, not about software. Every tool, from a phone app to a generator to a full editor, asks you the same seven things in some order, and the thumbnails that fail usually fail because one of the seven was never decided: the words were typed before the subject was chosen, or the shape was left to the export button. This post is the seven, in the order that makes each one easier.
It applies to a YouTube video, a Short, a podcast episode or a live stream, because the decisions are the same and only the frame changes. Where YouTube has said something in its own words, it is quoted with the date. The last section takes the same seven decisions through WThumb, where a reference stands in for the first five and you make the last two. If the term is new to you, start with what a YouTube thumbnail is.
The tool changes how each decision is carried out, not which decisions there are, and the main routes have guides of their own: how to make a YouTube thumbnail with AI from a link, how to make a YouTube thumbnail on your phone with no laptop, and how to make a YouTube thumbnail without Photoshop.
Decision one: the shape
Decide the frame before anything is placed in it, because the frame decides where things can go. As of September 2026, YouTube's Add custom thumbnails page asks for a 16:9 image for a video and 9:16 for a Short, names 3840 by 2160 and 2160 by 3840 as the resolutions, and sets 640 pixels as the minimum width. JPG or PNG, and a file size cap of 2 MB from the mobile app or 50 MB from desktop.
The shape is the decision people skip, and it is why a landscape thumbnail turns up on a Short as a strip with bars, or a square export from a phone app arrives with its edges cut. If the same video goes out in both shapes, it needs two thumbnails composed separately; a 9:16 crop of a 16:9 image keeps under a third of it and usually loses the words or the face.
Decide it, write it down, and make everything after it inside that frame. The exact numbers are in the size and resolution reference; for the rest of this post they are settled. In a layers editor, that means setting the canvas before anything else, the first step in both a GIMP YouTube thumbnail and a Photopea YouTube thumbnail.
Decision two: the one subject
A thumbnail is seen at about 210 pixels wide beside a title, on a phone, for a fraction of a second. At that size it can carry one subject. Not the product and the logo and the before and the after: one thing that tells a stranger what the video is about before the title is read.
A laptop split in half says a durability test. A pan on fire says a cooking disaster. A number that is absurdly large says a challenge. The cooking YouTube thumbnail ideas sort a kitchen's subjects into five jobs, and the home renovation thumbnail ideas show why the finished room beats the before and the after side by side.
The test for the subject is whether it survives losing the title. Cover the title with your thumb and look at the tile. If you cannot guess the premise, the subject is wrong, or the video has no premise simple enough to draw, which is a script problem the thumbnail cannot fix. As of September 2026, YouTube's Thumbnail & title tips page puts the same idea as a warning: "Try not to make the design too complex. Dynamic use of color and composition can help catch the eye, but too much can overwhelm it."
Choose the subject before opening any tool. It is the decision that costs nothing to change now and everything to change later, because the face, the words and the color are all placed in relation to it.
Decision three: a face, or none
A face gives the viewer someone to read the premise through. It works when the expression is a reaction to the subject in the same frame, so the two make a small story: this person saw that thing and felt this. The same YouTube tips page says content for casual viewers "might focus on actions and emotions that are more universally relatable, like a shocked face." Shot from the front in daylight, close, with the expression bigger than feels natural, because it will be shrunk.
No face is a real choice, not a fallback. Product channels, tutorials, screen recordings, music and many story channels do better with the subject filling the frame and a hand, an arrow or a split standing in for the reaction. What is never a choice is someone else's face: a real person's likeness in your thumbnail is theirs, and a tool that pastes one in from a reference is a problem before the design is even judged. Cover songs are the music exception: the cover song thumbnail post puts the performer's own face first.
Decide face or no face for the channel, not per video, so the tiles read as one channel in a subscription feed. The photo choice for an AI thumbnail covers how to shoot the photo when the answer is yes.
Decision four: the words, two or three at most
The words on a thumbnail are not the title. The title says what happens; the thumbnail words say the one stake, number or contrast that makes it worth clicking, and they must be readable at 210 pixels on a phone held at arm's length. Two words, three at the outside, in capitals: WORTH IT?, 24 HOURS, $1 VS $100. If the subject is unmistakable, no words is a valid answer.
Heavy plain typeface, a thick outline or a solid block behind the letters, and never over the busy part of the picture. YouTube's tips page: "If you add text, make sure to use a font that's easy to read." Thin weights, script fonts and long phrases fail on a phone before they fail anywhere else. Where the words are laid over a generated picture, as on the overlay route in CapCut's AI thumbnail maker, pick the corner and the color after the picture exists.
Keep the words away from the bottom right corner, where YouTube prints the video's length over the picture, and a little in from every edge, because the corners are rounded on most surfaces. A chat assistant will not keep that corner clear unless you tell it, one reason the answer to can ChatGPT make YouTube thumbnails is a qualified yes.
Decision five: color that separates
Color on a thumbnail has one job: separating the subject, the face and the words from the background at small size. Two dominant colors and one accent, the accent kept for the subject or the words. A warm face against a cool background, a bright object against a dark field, a white edge around anything that would otherwise blend. If the channel has a brand color, it becomes one of the two.
Contrast includes light. The subject is lit from the front so there are no deep shadows across it, and the background is darkened or blurred so it has no detail to compete. Where the real scene was cluttered, the clutter goes. A thumbnail that looks slightly too loud on a monitor is usually right on a phone; a thumbnail that looks tasteful on a monitor usually disappears in a feed.
This is the last of the five design decisions. Everything from here is checking and shipping.
Decision six: the check at phone size
Before uploading, look at the thumbnail the way a viewer will: small, beside a title, next to other videos. Shrink it to about a fifth of your screen width, or put it in a mock feed, and ask four questions. Can you name the subject? Can you read every word? Does the face read as a reaction? Is anything important in the bottom right corner?
Then the truth check. YouTube's Thumbnail & title tips page shows thumbnails and titles "working together to create interest and tell a story", and the story has to be in the video. A thumbnail that promises an event the video does not contain is a policy matter as well as a taste one, and it costs the channel the viewers who click and leave.
If the words were generated rather than typed, read every letter and every digit, because a generated image can misspell a word or drop a digit, and the text check for an AI thumbnail is the one step no tool does for you.
Decision seven: the upload, and the swap
Export in the shape decided in step one, as a JPG for anything photographic or a PNG for flat graphics, under the file limit for where you upload from. Custom thumbnails need a verified account; if the option is missing in YouTube Studio, verification is the first thing to check. The clicks on a computer and on a phone are in how to change a YouTube thumbnail.
The upload is not the end of the decision. The same tips page recommends "experimenting with updating older thumbnails to increase your video's appeal to new viewers", so a thumbnail that is not working can be replaced. Keep the seven decisions written down per video; a swap is usually one decision changed, not seven, and knowing which one is what makes the second version better than the first.
How to do this in WThumb
In WThumb a reference carries the first five decisions and you make the last two. You bring a thumbnail whose shape, subject treatment, face, words and color already work, and the result is a new design built from that idea around your own video. The run:
- Pick the shape. YouTube 16:9, Shorts 9:16, or both from one brief. Each is composed for its own frame rather than cropped from the other.
- Bring the reference. Paste a public YouTube link, or upload an image. Choose it for the decisions it gets right for your premise, not for the channel it came from.
- Decide the face. Add your own photo and you are placed in the design; skip it and the person in the reference is removed and the space is rebuilt around the subject. Nobody else's face is ever placed.
- Type the words. Up to three lines, in capitals, or choose no headline. The title stays in the title.
- Name what must be true. One line of instructions: the subject is my actual product, attached as @1; the background is a kitchen. Anything not named is designed for you. The guide to describing a thumbnail gives the order to write it in.
- Run, then check. About thirty seconds later, do decision six on the result: shrink it, read the words, confirm the subject. Download the Full HD+ JPEG when it passes.
- Swap when needed. Describe one change to a finished thumbnail and get a new version, which costs a credit (1 credit = 1 thumbnail); a wrong subject is a new run from a sharper instruction.
Start from a simple reference
For the first thumbnail, choose a reference with one obvious subject and a short headline area, so it is easy to tell whether the brief worked. The link route is the quickest way in when the reference is a video you admire. To take the seven decisions through it on your own video, new accounts get two free credits, no card needed, and paid plans are there when you need more.
