HackerTube.dev

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

3Blue1Brown · 2.2M views · 2025-07-25 · 37:20

Open on YouTube ↗

If playback is restricted in your region, try opening the original video. Use Mark watched to add it to your marked videos.

Overview

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.

On Hacker News

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

Recommended by 1 HN people across 1 months · HN mentions 2025–2026

Topics

LLMs & generative AIMachine learning & deep learning
Search by tags
diffusionclipddpmclassifier-free guidanceimage generation

What people said

dilyevsky · 2026-06 · Recommendation

> 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]

Read the original HN comment ↗
pornel · 2025-09 · Reference

Breakthrough in image generation speed literally came from applying better differential equations for diffusion taken from statistical mechanics physics papers: [VIDEO]

Read the original HN comment ↗
elliotto · 2025-08 · Reference

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 migration. The improved quality of the new models outweighs any issues around prompting.

Read the original HN comment ↗

Sources

Video by 3Blue1Brown, played through YouTube.

Data and selection notes

HN summaries and labels may contain AI errors. HN dates refer to discussions, not video publication.

HN Library: HackerTube export (2026-10-31) and open-index/hacker-news. Upstream summaries and classifications may contain AI errors.

Views, duration and playback status refresh separately from the published catalog. When an update fails, existing data remains visible.

Selection checks
  • SaaS & AI candidates: Matches the configured criteria
  • Editor-approved picks: Editorial review is required

Rule matches identify candidates. They do not guarantee video quality or financial results.

About the library and its sources