Skip to content
HomeVideos
AllLongsCategoriesPlaylists
BooksBlogsPodcasts
HistoryListsWatch laterLiked videosFollowing

The videos, books, blogs and podcasts Hacker News keeps recommending, all in one place.

VideosBooksBlogsPodcastsAboutPrivacySettings

Data from Hacker News, HackerTube, HackerBooks, HackerBlogs and Open Library. Credits · Not affiliated with Y Combinator.

HomeVideosBooksBlogsPodcasts
Ctrl K

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

    3Blue1Brown

    1 HN recommender · recommended in 3 months, 2025–2026

    YouTube

    37:20 · 2.2M views · Jul 25, 2025

    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.

    AI, Machine Learning & DataLLMs & generative AIMachine learning & deep learning
    1. 27:14
      Transformers, the tech behind LLMs | Deep Learning Chapter 53Blue1Brown12M views • 2 years ago
    2. 26:10
      Attention in transformers, step-by-step | Deep Learning Chapter 63Blue1Brown4.9M views • 2 years ago
    3. 8:25
      Large Language Models from scratchGraphics in 5 Minutes375K views • 4 years ago
    4. 28:48
      LSTM is dead. Long Live Transformers!Seattle Applied Deep Learning541K views • 6 years ago
    5. 3:31:24
      Deep Dive into LLMs like ChatGPTAndrej Karpathy10M views • 1 year ago
    6. 1:56:20
      Let's build GPT: from scratch, in code, spelled out.Andrej Karpathy8.2M views • 3 years ago
    7. 3:07
      Neural Networks Demystified [Part 1: Data and Architecture]Welch Labs839K views • 12 years ago
    8. 20:39
      AI Language Models & Transformers - ComputerphileComputerphile354K views • 7 years ago
    9. 1:31:13
      A Hackers' Guide to Language ModelsJeremy Howard580K views • 3 years ago
    10. 42:40
      State of GPT | BRK216HFSMicrosoft Developer786K views • 3 years ago
    11. 27:07
      Attention Is All You NeedYannic Kilcher814K views • 8 years ago
    12. 59:48
      [1hr Talk] Intro to Large Language ModelsAndrej Karpathy4.3M views • 2 years ago
    13. 1:22:38
      CS480/680 Lecture 19: Attention and Transformer NetworksPascal Poupart374K views • 7 years ago
    14. 23:02
      Did We Just Change Animation Forever?Corridor Crew3.5M views • 3 years ago
    15. 39:43
      Do we need Attention? - Linear RNNs and State Space Models (SSMs) for NLPSasha Rush35K views • 3 years ago
    16. 57:57
      Lecture 1 | Introduction to Convolutional Neural Networks for Visual RecognitionStanford University School of Engineering2.9M views • 9 years ago
    17. 4:17
      AI Learns Noise Filtering For Photorealistic Videos | Two Minute Papers #215Two Minute Papers65K views • 8 years ago
    18. 2:13:35
      Let's build the GPT TokenizerAndrej Karpathy1.3M views • 2 years ago
    19. 24:07
      AI can't cross this line and we don't know why.Welch Labs2.5M views • 2 years ago
    20. 1:26
      Fake Obama created using AI video tool - BBC NewsBBC News1.7M views • 9 years ago

    3 comments from Hacker News

    1. @dilyevskyJun 2026· in “Playing with Vision Embeddings”

      > 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

      RecommendsView on HN
    2. @elliottoAug 2025· in “Training language models to be warm and empathetic makes them less reliable”

      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.

      CitesView on HN
    3. @pornelSep 2025· in “The maths you need to start understanding LLMs”

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

      CitesView on HN