Introduction
What are SREFs?
SREF stands for Style Reference. It's a way to control the style of generated images.
Understanding SREFs
The Concept of Style Space
-Imagine a vast, multi-dimensional space (let's say 100 dimensions).
-This space contains every possible style for images.
-Each point in this space represents a unique style.
How SREF Random Works
-SREF random picks a random point in this style space.
-It's like throwing a dart blindfolded into a 100-dimensional 🎯dartboard.
Random Number Generation
The computer uses a random number generator to choose this point. For example, if the space had only 3 dimensions, it might pick numbers like (5, 100, 32).
Applying the Style
Once a random point is selected, the AI applies this style to the image it's generating
In other words
-Think of it like a massive library of painting styles.
-SREF random is like blindly picking a book from the library.
-The sref number is like remembering which shelf you picked from, so you can find the same book again later.
Conclusion
By using SREF random, you're essentially telling the AI, "Surprise me with a style!" Each time you use it, you get a different, random style for your image.
How to Use This Site
Search by SREF Number
Enter a number to see SREF results from that number onwards.
Search by Keyword/Emotion
You can click on any of the listed keywords/emotions to see similarly tagged SREFS.
Search by Description
Type anything in the search bar, to get results that most closely match from the LLM created descriptions.
Search by Color
Find the SREFS that most closely include the color you select from the color picker.
Search by Style Similarity
Another way to find similar SREFS is by searching by style similarity. (WIP)
Search by Latest
See the latest SREFs added to the site.
Share your favorites
Copy the MJ or Niji share link to share your favorite SREFs with others.
Please Note
All SREF descriptions were generated by LLM. I cannot guarantee their accuracy or quality. For this first attempt at comprehensive style organization, the emphasis was on keeping costs low, and exploring the style data itself. So consider this a work in progress, likely to change and evolve over time.