Comparing DDPM sampling with and without noise in Stable Diffusion 2
Stable DiffusionDemo summary
The demo shows the difference in output quality when generating a 'tree in the desert' using Stable Diffusion 2, contrasting a sharp image with a blurry one when the noise step is removed.
Step-by-step
- Select the DDPM sampling method
- Enter the prompt 'tree in the desert'
- Ensure the line of code that adds noise at each step is active
- Generate the image
Options
- Remove the noise addition line of code to see the impact on image clarity
Watch out for
- Removing the noise addition step results in a tiny, sad, blurry output
Highlights
“we can get some really nice images”
All demos from “But how do AI images and videos actually work? | Guest video by Welch Labs”
- 10:170:20Comparing DDPM sampling with and without noise in Stable Diffusion 2CurrentThe demo shows the difference in output quality when generating a 'tree in the desert' using Stable Diffusion 2, contrasting a sharp image with a blurry one when the noise step is removed.Stable Diffusion· AI Image Generator
- 33:080:29Adjusting Classifier-Free Guidance scale in Stable DiffusionThe demonstration shows how increasing the guidance scale alpha from 0 to higher values causes a tree to physically grow and gain detail within a desert scene in Stable Diffusion.Stable Diffusion· AI Image Generator
- Watch “But how do AI images and videos actually work? | Guest video by Welch Labs” →
AI Image Generator
- 10:170:20Comparing DDPM sampling with and without noise in Stable Diffusion 2CurrentThe demo shows the difference in output quality when generating a 'tree in the desert' using Stable Diffusion 2, contrasting a sharp image with a blurry one when the noise step is removed.3Blue1Brown
- 33:080:29Adjusting Classifier-Free Guidance scale in Stable DiffusionThe demonstration shows how increasing the guidance scale alpha from 0 to higher values causes a tree to physically grow and gain detail within a desert scene in Stable Diffusion.3Blue1Brown