But how do AI images and videos actually work? | Guest video by Welch Labs

Channel:3Blue1Brown · Length:37:20 · Published:2025-07-25 · Views:2267527

Watch on YouTube

Summary

A guest video by Welch Labs on 3Blue1Brown explaining how text-to-image and video models work. It covers CLIP, shared embedding spaces, diffusion models (DDPM, DDIM), conditioning, guidance and negative prompts.

Why HN recommends it

HN uses it as a clear explainer for diffusion mechanics, such as negative prompts and faster sampling, and for joint text-image encoders.

WL & 3b1b made a nice video on it — HN, 2026

Machine learning & deep learning · LLMs & generative AI

Subtitles (excerpt)

Over the last few years, AI systems have become astonishingly good at turning text props into videos. At the core of how these models operate is a deep connection to physics. This generation of image and video models works using a process known as diffusion, which is remarkably equivalent to the Brownian motion we see as particles diffuse, but with time run backwards, and in high-dimensional space. As we'll see, this connection to physics is much more than a curiosity. We get real algorithms out of the physics that we can use to generate images and videos. And this perspective will also give us some really nice intuitions for how these models work in practice. But before we dive into this connection, let's get hands-on with a real diffusion model. While the best models are closed source, there are some compelling open source models. This video of an astronaut was generated by an open source model called WAN 2.1. We can add to our prompt and have our astronaut hold a flag, hold a laptop, or hold a meeting. If we cut down our prompt to just an astronaut, we get this. And if we cut down our prompt to nothing, we interestingly still get this video of a woman. If we dig into our WAN model's source code, we'll find that the video

On Hacker News

> Because you somehow need a giant training set which describes images in natural language, no? That's definitely one way - they train a text encoder together with an image encoder on a labelled set of images. WL & 3b1b made a nice video on it: [VIDEO] — dilyevsky, 2026-06
Breakthrough in image generation speed literally came from applying better differential equations for diffusion taken from statistical mechanics physics papers: [VIDEO] — pornel, 2025-09
My favourite instantiation of this weird invocation is from this AI video generator, where they literally subtract the prompt for 'low quality video' from the input, and it improves the quality. [VIDEO] I've just migrated my AI product to a different underlying model and had to redo a few of the prompts that the new model was interpreting differently. It's not obseleted, just requires a bit of mi… — elliotto, 2025-08

Related videos