LoRAs

What are LoRAs? Without going into details, they are trainable addons for image generation models that allow for greater knowledge of their subject matter and the mimicking of styles to degrees previously unachievable by embeddings or hypernetworks.

You can use them by installing the "sd-webui-additional-networks" extension available in the Extensions tab of your local image generation installation, putting the .safetensors files in the extensions/sd-webui-additional-networks/models/lora folder (or other one specified in the Settings tab), and applying them in their drop-down menu. You can mix LoRAs and achieve, for example, a character from one LoRA in the style from another.

Below is a collection of LoRAs trained on /aids/. All of them were trained on a mix of cleaned SFW and NSFW data for the animefull-final model.