308 - How Image Diffusion Models Work - the 20 minute explainer

You already know how LLMs work from our popular 20-minute explainer. Now we take it to images. What does Michelangelo have to do with stable diffusion? More than you'd think. Walk away knowing how image generation actually works — and what it has in common with the text models you already understand.

Full shownotes at fragmentedpodcast.com (fragmentedpodcast.com/episodes/308).

Show Notes

• Episode 303 - How LLMs work in 20 minutes - text generation

• VAE -
Variational Autoencoder (ibm.com/think/topics/variational-autoencoder)

• RGB Color model (en.wikipedia.org/wiki/RGB_color_model) - wikipedia

• Word2Vec technique (en.wikipedia.org/wiki/Word2vec) - wikipedia

• Efficient Estimation of Word Representation (arxiv.org/abs/1301.3781) -

original Word2Vec paper by Mikolov et al.

• High-Resolution Image Synthesis with Latent Diffusion Models (arxiv.org/abs/2112.10752) -

Rombach et al. (2022) — the paper behind Stable Diffusion

• Image Training data

• LAION-5B (arxiv.org/abs/2210.08402) - 5 billion image-text pairs

scraped from the web, used to train many image generation models

• WebLI (aibase.com/news/15390) - Google's internal image-text

dataset

• Michelangelo (goodreads.com/quotes/1191114-the-sculpture-is-…)

Get in touch

We'd love to hear from you. Email is the

best way to reach us or you can check our contact page for other

ways.

We want to hear all the feedback: what's working, what's not, topics you'd like

to hear more on.

• Contact us (fragmentedpodcast.com/contact)

• Newsletter (buttondown.com/fragmentedcast)

• Youtube (youtube.com/@fragmentedpodcast)

• Website (fragmentedpodcast.com)

Co-hosts:

• Kaushik Gopal (kau.sh)

• Iury Souza (iurysouza.dev)

[!fyi] We transitioned from Android development to AI starting with
Ep. #300. Listen to that episode for the full story behind

our new direction.