How AI Image Generators Work (Diffusion, Visualized)

1 Просмотры

  • Спасибо! Поделитесь с друзьями!

    Вам не понравилось видео. Спасибо за то что поделились своим мнением!

aii
Издатель
How AI image generators actually work, explained visually: they do not paint, they carve a picture out of pure random noise, one denoising step at a time.

A full, visual explanation of how diffusion models turn a text prompt into an image. We show why generation is really denoising run in reverse, how the model is trained by first adding noise to real images, why the whole process runs in a compressed latent space instead of on pixels (the trick that made it cheap), how a U-Net or transformer predicts the noise, and how your words steer every step through cross-attention. We finish with seeds, schedulers and guidance, and why extra fingers and gibberish text are the honest fingerprint of a system built on local plausibility. Built from the public research on latent diffusion models.

Chapters:
0:00 A picture out of noise
4:06 Learning to remove noise
9:00 The trick: a smaller space
13:34 How your words steer it
18:16 Why it fails the way it does
22:39 A very good guess

If you enjoyed this video, consider subscribing: https://www.youtube.com/@codetechnologyy?sub_confirmation=1

#AI #Diffusion #StableDiffusion #MachineLearning #GenerativeAI
Категория
AI Картинки
Выполните вход или зарегистрируйтесь для отправки комментария.
Комментариев нет.