But how do AI images and videos actually work? | Guest video by Welch Labs
3Blue1Brown1 HN recommender · recommended in 1 month, 2025–2026
explainer · intermediate · evergreen · 2025–2026 on HN
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.
HN uses it as a clear explainer for diffusion mechanics, such as negative prompts and faster sampling, and for joint text-image encoders.
Up next
10 mo on HNTransformers, the tech behind LLMs | Deep Learning Chapter 5
3Blue1Brown
13 HN recommender · 2024–2026
8 mo on HNDeep Dive into LLMs like ChatGPT
Andrej Karpathy
10 HN recommender · 2025–2026
8 mo on HNLet's build GPT: from scratch, in code, spelled out.
Andrej Karpathy
14 HN recommender · 2023–2025
8 mo on HNA Hackers' Guide to Language Models
Jeremy Howard
8 HN recommender · 2023–2024
6 mo on HN[1hr Talk] Intro to Large Language Models
Andrej Karpathy
9 HN recommender · 2023–2026
4 mo on HNSparks of AGI: early experiments with GPT-4
Sebastien Bubeck
7 HN recommender · 2023–2026
5 mo on HNLet's build the GPT Tokenizer
Andrej Karpathy
5 HN recommender · 2024–2025
4 mo on HNAttention in transformers, step-by-step | Deep Learning Chapter 6
3Blue1Brown
5 HN recommender · 2024–2026
4 mo on HNThe A.I. Dilemma - March 9, 2023
Center for Humane Technology
11 HN recommender · 2023–2024
4 mo on HNLSTM is dead. Long Live Transformers!
Seattle Applied Deep Learning
4 HN recommender · 2020–2023
4 mo on HNAttention Is All You Need
Yannic Kilcher
4 HN recommender · 2020–2025
4 mo on HNPlaying a Neural Network's version of GTA V: GAN Theft Auto
sentdex
3 HN recommender · 2021–2025
3 comments from 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: ▶ this video
Breakthrough in image generation speed literally came from applying better differential equations for diffusion taken from statistical mechanics physics papers: ▶ this video
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. ▶ this 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.