diffusion-xl / DOCS.md
adamelliotfields's picture
Sync with adamelliotfields/diffusion
0d34381 verified
|
raw
history blame
2.29 kB

Usage

TL;DR: Enter a prompt or roll the 🎲 and press Generate.

Prompting

Positive and negative prompts are embedded by Compel for weighting. See syntax features to learn more.

Use + or - to increase the weight of a token. The weight grows exponentially when chained. For example, blue+ means 1.1x more attention is given to blue, while blue++ means 1.1^2 more, and so on. The same applies to -.

For groups of tokens, wrap them in parentheses and multiply by a float between 0 and 2. For example, a (birthday cake)1.3 on a table will increase the weight of both birthday and cake by 1.3x. This also means the entire scene will be more birthday-like, not just the cake. To counteract this, you can use - inside the parentheses on specific tokens, e.g., a (birthday-- cake)1.3, to reduce the birthday aspect.

Models

Each model checkpoint has a different aesthetic:

Scale

Rescale up to 4x using Real-ESRGAN with weights from ai-forever. Necessary for high-resolution images.

Advanced

DeepCache

DeepCache caches lower UNet layers and reuses them every n steps. Trade quality for speed:

  • 1: no caching (default)
  • 2: more quality
  • 3: balanced
  • 4: more speed

Refiner

Use the ensemble of expert denoisers technique, where the first 80% of timesteps are denoised by the base model and the remaining 80% by the refiner. Not available with image-to-image pipelines.