I tested the first major AI image generator, Dall-E, when it originally launched. Since then, I’ve watched as the world of generative AI has exploded, but one feature has always bugged me: text in images.

As faces looked clearer, and hands went down to the correct number of fingers, every model still seemed to really struggle with creating text.

With each major upgrade to different AI image generators, both those built into chatbots like Gemini or ChatGPT, as well as standalone models like Midjourney, I have seen huge progress being made.

But as faces looked clearer, and hands went down to the correct number of fingers, every model still seemed to really struggle with creating text. Whether that was on a poster, a sign or even a T-shirt, it often looked like a giant smudge of hieroglyph

See Full Page