The new GPT-4o image generator is seriously impressive. The jump in quality (vs. the previous DALL·E 3 image generation model) is obvious right away. It handles people so much better than before. Faces look realistic, proportions make sense, and even details like posture and lighting often feel intentional. It’s not just spitting out random visual noise anymore. It actually looks like it knows what it's doing.

One thing that really stood out is how well it generates text. I asked for posters, signs, even labeled diagrams, and most of the time the writing was perfectly readable. That alone makes it way more useful than older versions. It also does a much better job with character consistency, which is something other tools like Midjourney still struggle with.

That said, it’s not flawless. It still struggles with certain prompts, especially fantasy-related ones like maps or magical creatures. Some filters are still pretty strict. Also, the rollout is kind of all over the place. I got it on my phone but not on desktop at first. But when it works, it’s genuinely exciting. I’ve already started using it for mockups and visuals I used to need Photoshop for.