Prompting With SDXL

Modified on Thu, 28 Sep 2023 at 02:25 AM

Prompting with SDXL based models is slightly different than prompting with SD1.5 or SD2 based models.

SDXL relies on two CLIP models, rather the singular CLIP model used by SD1.5 and SD2 models. CLIP models

This means prompts for SDXL have two components for the positive and negative text inputs:

  • Prompt is the prompt describing the subject and composition of the image
  • Style is the artistic style in which the generated image should created

Concatenate Prompt & Style

Along with two prompt components, prompt & style can be concatenated. This means that the style prompt is appended to the prompt and this new concatenated prompt is sent to both CLIP models.

When starting out with SDXL, it is recommended to concatenate prompt & style.

Recommended Usage

  • Prompt: [key elements of the prompt, including subject, aesthetic terms, etc.]
  • Style: [any other stylistic additions to the prompt]
  • Enable Concatenation (the link between prompt & style)

Experiment by turning Random Seed off and exploring what changing the prompt does to the same image generation process.


Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select atleast one of the reasons

Feedback sent

We appreciate your effort and will try to fix the article