Garbled text in AI images has been the dead giveaway for two years, so a model that finally spells properly and keeps a subject consistent across 14 reference shots is worth a test run on your next thumbnail. I used to use Nano Banana, but then OpenAI updated their image generation, and I found it quite good.
I will definitely now balance between Nano Banana and ChatGPT once the new version of Nano Banana is available. I have always found image generation to be hit or miss. Sometimes it is amazing, and other times I could probably do it better myself (actually, I probably couldn’t, but you get the idea).
Any kind of generative AI (text, video, image, or code) is only as good as the context you provide. The initial prompt truly sets the stage. What will be interesting to test is how well you can guide AI after the initial generation and how well it follows your instructions.