Improvements to the inbuilt image generation

I love that Base44 has inbuilt AI image generation but I noticed this week that the quality is behind what you see in ChatGPT (OpenAI), even though I understand they are using the same model under the hood, not sure why.

Either way, if it’s using something else, it needs to be improved so that image generation is of higher quality and doesn’t for example make typos when adding text to the images or adds extra/unnecessary texts. You would see these type of mistakes all the time in the older AI image generation models, but you don’t see them today. Those issues are pretty much gone from ChatGPT, Ideogram, etc. the image generation nowadays is pretty much flawless and of incredible quality, but the AI image generation inbuilt into Base44 seems to be using an older model since I’m seeing the errors I would see in those older models.