Preview iImages https://imgur.com/a/8d0JLcA
IMPORTANT INSTRUCTIONS!!!
This model was trained on SD base 1.5 version BUT It does also work for 1.4 as they both share the same Clip encoder.
Install instructions.
Simply place the water elemental.pt file inside the \stable-diffusion-webui\models\hypernetworks folder. Load the model inside the Automatic1111 interface under settings hypernetwork.
Use instructions.
Use between 0.55-1.0 hypernetwork strength, more strength will give a more transparent elemental look but starts to overfit. I find .7 works well enough.
Use DPM++ SDE Karras sampler with 15 steps and CFG of 7.0.
Make sure and always include the word water elemental somewhere in the prompt. For people always preface the subject with water elemental, example "water elemental man walking", "water elemental girl playing in the backyard", etc...
VERY IMPORTANT! Always describe the background in some detail or you WILL get a very generic boring background.. So for example DON'T just say "an old water elemental man". DO say "an old water elemental man inside a rustic hut".
Some fun info. People have been sleeping on hypernetworks and I plan to change that. Hopefully the flexibility of this hypernetwok will show everyone their true potential. Because this model is a hypernetwork it can be used in conjunction with ANY model based on the 1.4 CLIP architecture. That means this model will work on any custom 1.4 or 1.5 model, like the modern disney model, or classic disney, etc… for example, let's say you want to load classic disney as base. Well simply load the classic disney model, make sure and preface every prompt with classic disney. As per instructions of the model. Then follow up with my “water elemental” tag as instructed once you have loaded the hypernetwork. So the prompt should look something like this “classic disney. water elemental girl playing in the backyard.” Have fun folks!
Also feel free to check out my basic starter guide on training your own models. https://pdfhost.io/v/SnKTqK5ca_Untitled_document