There is a LinkedIn user who writes and titles posts on various technical topics…and then ruins the posts by illustrating them with an AI image generator and including the post title in the image.
Why are they ruined? Because some I mage generators can’t spell. For example, one of the images includes the so-called word CYBERCEERITYY.
If you want your AI-generated images to include correctly-spelled words, either use short simple words or use large multimodal model (LMM) image-text generators. See my earlier post, “Text Generation in Images? Use an LMM” at the https://bredemarket.com/2024/07/30/text-generation-in-images-use-an-lmm/ URL.
Apparently Google Gemini’s Imagen 4 incorporates LMM capabilities; I fed it a complex textual title and it was spelled correctly in the resulting image.
(Imagen 4)
Bredemarket 