There is no denying that noise diffusing technology can create some impressive looking imagineries

It’s quite comfy for many YouTubers to just type any random garbage they are currently making video about into the dalle-3 prompt and call it a day. Unfortunately it was never ought to be that simple. Thumbnail creation comes with rather script rules and in punishes it greatly for not following them, yet even tho many channels using it show great decline in views since start of this process, they are keep doing it.

Issues

So why exactly this AI generated thumbnail doesn’t work? Well there are a couple of reasons.

1. Color spectrum The noise diffusers suffer from some weird illness that’s make them handle color and contrast diversity not all that good. It’s not all that clear on the beginning, but for someone who have been working with them since beginning is just one sight to notice this peculiar characteristic. There are some fixes for that contrast-fix, better-colors, but the truth is that it does requite major color work, to be called a good image.
2. Composition The positioning of stuff on the plane is quite important as well, thumbnails need to catch attention instantly, so the placement of given items need to be precisely deliberate as well.
3. Emotions If someone looks at your video miniature and doesn’t feel anything then it means that it’s not particuallry good. Achieving this with AI generated images is rather hard, not immpossible, but usually harder than doing it manually. The prompt can yet not interprete thing like “great miniature”, “good converting thumbnail” or “attention catching”, it will do something, but with reasonably unimpressive effect.
4. Informative Noise diffusers can create good-looking images, but not necessarily precisely informative, and since that’s the main goal of thumbnail, then it speaks for itself…

So should using AI generated art for YouTube thumbnail should be avoided for any cost? No it’s a tool, and it needs to be used right, otherwise it might be harmful.