So you think you know Text to Video Diffusion models?

Описание: Video Diffusion Generative AI is the next frontier for AI. In this video we discuss the problem, the challenges, the solutions, and the seminal papers in the field like Google's Imagen, Meta's Make-a-video, Nvidia's Video Latent Diffusion Model (LDM), and OpenAI's SORA. On the way, we discuss the core concepts of Image Diffusion models, like Forward and Reverse Diffusion, UNet, convolution, and diffusion transformers. This video is meant to be a quick overview of all the major concepts in the field - hope you guys and gals found it useful for deeper dives.

Buy me a coffee at https://ko-fi.com/neuralavb !
Support us on Patreon to access slides and video material!
patreon.com/NeuralBreakdownwithAVB

Related videos:
What are Conditional Image Diffusion Models?
   • Text to Image Diffusion AI Model from scra...  

What is Latent Space?
   • Visualizing the Latent Space: This video w...  

How do LLMs generate images? (The answer is not diffusion)
   • If LLMs are text models, how do they gener...  

Transformers and Attention Playlist
   • Everything Language Processing and LLMs!  

Visit our Patreon for full access to code and other documents/animations:
  / neuralbreakdownwithavb  

#generativeai #deeplearning #ai

Useful papers:
Video Diffusion Models: https://arxiv.org/abs/2204.03458
Imagen: https://imagen.research.google/video/
Make A Video: https://makeavideo.studio/
Video LDM: https://research.nvidia.com/labs/toro...
CogVideoX: https://arxiv.org/abs/2408.06072
OpenAI SORA article: https://openai.com/index/sora/
Useful article: https://lilianweng.github.io/posts/20...
Survey Papers: https://arxiv.org/abs/2310.10647 and https://arxiv.org/abs/2405.03150


Timestamps:
0:00 - Intro
0:39 - Text to Image Conditional Diffusion Models
2:16 - Challenges with Video Diffusion Models
3:43 - VDM (2022)
4:50 - Factorized 3D Unet models
5:46 - Meta Make A Video
7:28 - Google Imagen Video
8:07 - Nvidia Video LDM
9:36 - OpenAI SORA