AI YouTube thumbnail maker: covers people click, made from a brief
By VODOO AI · · 4 min read
On YouTube, the thumbnail decides whether anyone watches. It's the size of a stamp on a phone, it sits next to a dozen others, and it has about a second to win. The good ones follow a few simple rules, and AI makes it quick to try several versions that follow them.
This guide covers the size, what goes in the frame, how to get text that reads at stamp size, briefs to copy, and how YouTube lets you test two thumbnails on the same video.
The size
Shape
- Use
- 16:9, wide
Size
- Use
- 1280 by 720 pixels
File
- Use
- JPG or PNG, under 2 MB
Shorts
- Use
- 9:16 cover frame, chosen in the upload screen
| What | Use |
|---|---|
| Shape | 16:9, wide |
| Size | 1280 by 720 pixels |
| File | JPG or PNG, under 2 MB |
| Shorts | 9:16 cover frame, chosen in the upload screen |
One face, one object, three words
- One face, big, showing a clear feeling: surprise, delight, doubt.
- One object, the thing the video is about, large and sharp.
- Three words or fewer of text, never the video's title repeated.
- Strong contrast: a bright subject on a plain, darker background.
- Nothing small: if you can't make it out at stamp size, leave it out.
Look at your thumbnail on your phone, in the YouTube app, next to others. That's where people will see it, and it's where weak thumbnails show themselves.
Text that reads
Some engines are much better with lettering than others. Ideogram gets short words right, in bold shapes, even on busy backgrounds. Put the exact words in quotes in your brief and say where they go: "the words 'IT WORKED?' in huge white letters top left".
Briefs to copy
A YouTube thumbnail, 16:9: a woman in her twenties on the right, mouth open in surprise, holding up a tiny espresso machine on the left, bright yellow background, the words 'TINY BUT?' in huge bold black letters top left. Sharp, high contrast, nothing in the bottom-right corner.Make this in VODOO →
A YouTube thumbnail, 16:9, split down the middle: a messy, dark garage on the left, the same garage bright and organised on the right, a big red arrow between them, the words '1 WEEKEND' in huge white letters. Clean, punchy, high contrast.Make this in VODOO →
A YouTube thumbnail, 16:9: a man in his forties with glasses on the left, looking doubtfully at a glowing smartphone he holds up, a dark blue background, the words 'WORTH IT?' in huge yellow letters on the right. Sharp face, strong rim light.Make this in VODOO →
A look people recognise
Viewers who liked one video should spot the next one in their feed before they read a word. Keep two or three things the same on every thumbnail: the same background colour, the same lettering, the same side for the face. Write those into a short line you paste into every brief, such as "bright yellow background, bold black lettering top left, face on the right".
Test two versions
YouTube Studio can test up to three thumbnails on the same video and keep the one people watch most. Make two or three that differ in one thing, the face, the words or the colour, and let YouTube pick. Over a few videos you'll learn what your viewers stop for.
Keep the thumbnail honest. Showing something the video doesn't deliver gets clicks once, then people stop trusting the channel, and YouTube notices when viewers leave early. If the person in the thumbnail is made with AI and looks real, say so in the video.
Make your first thumbnail
- 01Open CreateChoose Photograph, set 16:9, and describe the face, the object and the words.
- 02Make threeAsk for three that differ in one thing.
- 03Test in YouTube StudioUpload them as a test on the same video and keep the winner.
Start in Create with 300 free credits a month, no card needed. For the video itself, see AI YouTube Shorts.
Questions people ask
- What size is a YouTube thumbnail?
- 1280 by 720 pixels, 16:9, as a JPG or PNG under 2 MB.
- Can AI put text on a thumbnail?
- Yes. Put the exact words in quotes and say where they go. Ideogram is the best engine for short, bold lettering.
- How many words should a thumbnail have?
- Three or fewer, and never the title repeated. The words should add something the title doesn't say.