What is a thumbnail generator?

A tool that makes the image YouTube shows before someone clicks a video. In ThumbnailMaker you paste the video's link or describe it in a sentence, keep or change the words and the style it proposes, and it generates a 16:9 thumbnail or a 9:16 Shorts cover with the words drawn into the image and read back.

Is there a free AI thumbnail generator?

Writing the brief from your video's link is free with an account; generating takes a plan. Free generators exist, mostly design tools with templates or an AI image with a daily limit, and the best AI thumbnail makers, compared says which ones are free and what each leaves to you.

Can it make thumbnails for YouTube Shorts?

Yes. Pick Shorts and the thumbnail is generated at 9:16, with the words kept clear of the title and the buttons the Shorts player lays over the bottom of the frame. The YouTube Shorts thumbnail maker page shows it on real outputs.

What size are the thumbnails?

1280 × 720 for videos and 1080 × 1920 for Shorts, the sizes YouTube recommends, with a 4K download as well. Every number is in the YouTube thumbnail size guide.

Can I change a thumbnail after it is generated?

Yes, in a sentence: "make the title yellow", "move her to the left", "remove the boats". The change is applied to that image and the rest of it stays as it is, so you don't start over to fix one detail.

Writing the brief from a video's link is free with an account. Generating, uploading and editing take a plan: Starter is $12 a month for 400 credits, and one thumbnail or one edit is 10 credits, so about 40 thumbnails a month.

AI YouTube thumbnail generator for any kind of video

Thumbnail Generator

ThumbnailMaker is a thumbnail generator that starts from your video, not from an empty prompt box. Paste the YouTube link: it writes the brief, generates the thumbnail and reads the words back before you see it.

Most thumbnail generators draw a picture. The thumbnail is the rest of the job. Type a prompt into an image model and you get an image: often a good one. Then the work starts. The title comes out misspelled, the face sits where YouTube's duration badge will cover it, the scene repeats the video's title instead of adding to it, and the 16:9 crop cuts the one thing that mattered. A thumbnail generator worth paying for does that part too: it decides what the image should promise from what the video is about, keeps the text to four or five words and spells them right, frames for 1280 × 720 from the start, and checks the result at the 336 px a viewer actually sees in the feed. That is what ThumbnailMaker does after you paste the link.

How it works

You bring the travel video. We make the thumbnail.

Here is one real thumbnail from this page, followed from the brief to the download.

Your input
Paste the video's link, or describe it
The brief is written from the link for you, or you write the scene yourself. This one was one sentence: “A young backpacker woman stands at the edge of a turquoise lagoon at sunrise, arms spread wide, a small worn backpack at her feet, limestone cliffs and long-tail boats behind her”
Words and style
The exact words, and a style
The words $30 A DAY, typed exactly as they should read.
Generated, then checked
Checked at the size of the feed
Every image is graded at 336 px, the width of a card in YouTube's feed. The words are read back and redrawn if a letter is wrong, and the bottom-right corner stays clear of the duration badge.
Your thumbnail
Download it, or change it in a sentence
1280 × 720, or 4K. If it is close, describe the change ("make the title yellow") and the rest of the image stays as it is.

AI-generated travel thumbnail: a backpacker with her arms spread wide at a turquoise lagoon at sunrise, with the words $30 A DAY — The result of step 04, as generated.

More from the same generator

AI-generated cooking thumbnail: a home cook lifts the lid off a pot of red curry in a warm kitchen, with the words ONE-POT CURRY — A recipe video

AI-generated renovation thumbnail: a woman in safety glasses holds a drill in a half-finished kitchen, with the words $500 KITCHEN — A renovation video

What it does that a prompt box doesn't

It starts from the video

Paste the link and the brief is written from the video itself: its title, its description and what is said in it. You get a scene, three to five words and a style to keep or change. No prompt to write, and no prompt-writing guide to read first.

youtu.be/… → "A backpacker at a turquoise lagoon at sunrise, arms spread wide" + $30 A DAY

The words come out spelled right

Image models still misspell. Every thumbnail's text is read back from the finished image and compared with what you asked for, and a wrong letter is redrawn before the image reaches you, not after you notice it on your channel.

Checked at the size of the feed

A thumbnail is seen at about 336 px wide, not at full screen. Each one is graded at that size: can the words be read, is there one clear subject, and is the bottom-right corner free for the duration badge.

Three versions to test, on Pro

On the Pro plan, Three to test makes three thumbnails from three different angles on the same video, ready for YouTube Studio's A/B Testing. The point is to learn which promise your viewers click, not to pick the prettiest image yourself.

Mistakes that make generated thumbnails look generated

The video's title, painted on

The title is already printed under the thumbnail. Repeating it in the image wastes the space and is unreadable on a phone. Three or four words the title doesn't say, like $30 A DAY on a video called Thailand on a Budget, do more.

The airbrushed stock face

Waxy skin, perfect teeth and an expression no one makes in real life read as AI from across the room. That is a prompting failure, not a limit of the model: the brief asks for a real face doing one real thing.

Five ideas in one frame

A face, a product, two arrows, a red circle and a logo: at 336 px that is noise. One subject and one promise per thumbnail, and the rest goes in the title.

Designing at one shape, cropping to another

An image made square or 3:2 and cropped to 16:9 loses its edges, and often the hand, the object or the word that carried it. Generate at the ratio YouTube shows, 16:9 or 9:16 for Shorts.