The thumbnail is the advert; the title is the caption underneath it. On a phone the image is roughly two centimetres wide, which means a thumbnail that works is one you can understand at a glance, with the sound off, while scrolling past it.
This tool returns five concepts rather than five images: a composition, a text overlay and a direction you can hand to a designer, shoot yourself, or rebuild in Canva in ten minutes.
One idea per thumbnail
The most common mistake is cramming three ideas into one frame — a face, a product, a location and four words of text. At thumbnail size that becomes grey mush. Pick the single thing that makes the video worth watching and let it occupy most of the frame.
A useful test: shrink your draft to 120 pixels wide and look at it. If you cannot tell what it is, neither can anyone else.
Text: three or four words, no more
Text on a thumbnail should not repeat the title — it should add the missing half. If the title says what the video is, the thumbnail text says why it matters. Three or four words is the practical ceiling at mobile size.
Heavy sans-serif, high contrast, and a stroke or shadow so it survives being placed over a busy image. Avoid the bottom-right corner: the duration badge sits there.
Contrast is what makes it visible
Your thumbnail is not seen alone, it is seen in a grid of eleven others. Whatever the dominant colour of your niche is — and most niches have one — moving away from it is worth more than any amount of polish. Faces with a clear, readable emotion still outperform objects in most categories, because people look at eyes first.
Make two and let the data pick
If your channel has access to A/B testing, produce two of the five concepts and run them. If it does not, publish one, note the click-through rate after 48 hours, and swap in the other if it sits under 4%. Thumbnails are the one asset you can change after publishing without penalty — use that.
Frequently asked questions
Does this generate the actual thumbnail image?
No. It generates concepts — composition, text overlay and art direction — that you can execute in Canva, Photoshop or Figma in a few minutes. Ideas are the slow part; drawing is not.
What size should a YouTube thumbnail be?
1280 by 720 pixels, 16:9, under 2 MB, in JPG, PNG or WEBP. Design it at that size but always check it shrunk to about 120 pixels wide, which is how most people will actually see it.
How many words of text should a thumbnail have?
Three or four. Beyond that the text becomes unreadable at mobile size, and it competes with the title rather than complementing it.
Is it free?
Yes, with no signup. Generate as many concepts as you like.
Do faces really perform better?
In most categories, yes, when the expression is legible at small size. The effect comes from readable emotion, not from the face itself — a flat expression performs no better than an object.