Currently I am using FreePik (paid not free version) and illustrator to make thumbnails.

I would love to instead use stablediffusion to generate thumbnails and in the future hero images

In particular, we like images that look more like scenes - the two on the left. There aren’t enough topical images like those two and I often have to use more generic clipart like the two on the right. I also spend a lot of time adding diversity to clip-art which presumably stable diffusion could do from the get-go.

Is anyone successfully using AI to generate their thumbnails? Happy to pay for a good tool!