← All notes

Summary

High-Resolution Image Synthesis with Latent Diffusion Models (2021) made high-resolution diffusion practical by moving the expensive generative process into a compressed latent space. That design is the basis of Stable Diffusion and of many later open image models.

What it introduced: latent diffusion — train a diffusion model in the latent space of a pretrained autoencoder, with optional text (or other) conditioning — so that high-resolution synthesis no longer requires operating in raw pixel space.

What it led to: Stable Diffusion and the open image-generation ecosystem that followed. Live notes: hf-diffusers-pipeline-loading and hf-diffusers-memory. Read it next to the denoising-diffusion pointer in this batch, which is the pixel-space predecessor.

Where to read it: https://arxiv.org/abs/2112.10752 (v2, last revised 13 Apr 2022). CVPR 2022.

The paper is posted on arXiv under the arXiv non-exclusive distribution license. This note links to it rather than reproducing it. The note itself is CC BY 4.0.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

@misc{iil_note_e53qbka9,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{High-Resolution Image Synthesis with Latent Diffusion Models}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2112-10752-latent-diffusion-models/},
  note = {Permanent ID iil:note:e53qbka9. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2112-10752-latent-diffusion-models.md.