Image generation AI has been good enough to convert images into Ghibli for awhile, but it took effort and a custom model/LoRA/pipeline being built.
GPT-4o image generation requires neither, which changes the generative image AI utility significantly. For better or for worse.
GPT-4o image generation requires neither, which changes the generative image AI utility significantly. For better or for worse.
Comments
We can *start* with AI slop that doesn't trigger the uncanny valley. From the demo examples, the image generations have better prompt adherence than anything else on the market for complex prompts.
It's far more accessible than other services/free, which has been the reason slop took off.