Update README.md

README.md

@@ -19,7 +19,7 @@ widget:
 
 Anything3.0 is an overfitted model that takes liberties when it shouldn't be generating human images and certain details. However, the community has given it a high rating, and I believe that is because many lazy people who don't know how to write a prompt can use this overfitted model to generate high-quality images even if their prompts are poorly written.
 
-Here is a ACertain version of Anything3.0, made with Dreambooth, initialized with [ACertainModel](https://huggingface.co/JosephusCheung/ACertainModel).
+Here is a ACertain version of Anything3.0, made with Dreambooth (idea of [LoRA](https://arxiv.org/abs/2106.09685) integrated), initialized with [ACertainModel](https://huggingface.co/JosephusCheung/ACertainModel).
 
 Although this model may produce better results for image generation, it is built on two major problems. Firstly, it does not always stay true to your prompts; it adds irrelevant details, and sometimes these details are highly homogenized. Secondly, it is an unstable, overfitted model, similar to Anything3.0, and is not suitable for any form of further training. As far as I know, Anything3.0 is obtained by merging several models in just the right way, but it is itself an overfitted model with defects in both its saturation and configuration. However, as I mentioned earlier, it can make even poorly written prompts produce good output images, which leads many lazy people who are incapable of writing good prompts to quickly surpass those who study the writing of prompts carefully. Despite these problems, I still want to release an extended version of the model that caters to the preferences of many people in the community. I hope would you like it.