Every AI image you've ever seen started as random TV static. Here's how diffusion models (the tech behind Stable Diffusion, Midjourney, DALL·E and Imagen) turn noise into a cat astronaut on the moon, one denoising step at a time.
During training, the AI watches millions of images get destroyed with noise. It learns only one skill: undo a single step. Then it starts from pure static and removes noise again and again, guided by your prompt, until a picture appears.
During training, the AI watches millions of images get destroyed with noise. It learns only one skill: undo a single step. Then it starts from pure static and removes noise again and again, guided by your prompt, until a picture appears.
- Категория
- AI Картинки
Выполните вход или зарегистрируйтесь для отправки комментария.
Комментариев нет.
