Girl's face destroyed

I tested the character consistency with the newest GPT4o image generator on sora. AI still needs some development and needs an open source alternative to be able to have more freedom, but for now the character consistency started to be good and some violence are also allowed by the "high and mighty" OpenAI content policy (lol) It have less random errors than earlier diffusion based generators and the requesters have much more control to the generated content and it allow more intelligent editing. Still not perfect but image generation with omni models are works mush better than diffusion based models. I am also not sure if its good or bad but the technology looks unstoppable.