Pinned Loading

1. taesd taesd Public

   Tiny AutoEncoder for Stable Diffusion (and other image models)

   Python 968 52

2. taehv taehv Public

   Tiny AutoEncoder for Hunyuan Video (and other video models)

   Python 522 21

3. maple-diffusion maple-diffusion Public

   Stable Diffusion inference on iOS / macOS using MPSGraph

   Swift 822 54

4. dino-diffusion dino-diffusion Public

   Bare-bones diffusion model code

   Jupyter Notebook 160 14

5. seraena seraena Public

   WIP Pytorch code for stably training single-step, mode-dropping, deterministic autoencoders

   Jupyter Notebook 50 6

6. notes_on_sd_vae notes_on_sd_vae

   1

   # Notes / Links about Stable Diffusion VAE

   2

   3

   Stable Diffusion's VAE is a neural network that encodes images into a compressed "latent" format and decodes them back. The encoder performs 48x lossy compression, and the decoder generates new detail to fill in the gaps.

   4

   5

   ![](https://user-images.githubusercontent.com/8733615/281488415-ea77f359-d380-412d-b27a-21884d19da1d.jpg)