Basic Introduction to Use Models in Stable Diffusion WebUI for Beginners

Along with the main model, there are few kinds of models we can use in Stable Diffusion Web UI to generate images.

Here I will introduce how to use these models in the Web UI.

Find Models You Need

There are community sites for people to share and download Stable Diffusion models.

Among them, civitai and Hugging Face are the most popular ones.

You can download models from these sites and use them in Stable Diffusion Web UI.

They also each have an extension for Stable Diffusion models.

• Official civitai extension for Stable Diffusion Web UI
• Hugging Face extension for Stable Diffusion Web UI by camenduru

If you are new to Stable Diffusion, before you start, I recommend you to browse gallery in community sites to find digests from other users' works along with the models and prompts they use.

• civitai Gallery
• Hugging Face Diffusers Gallery
• PromptHero
• OpenArt

Stable Diffusion Model

Stable Diffusion Model is the main model used in Stable Diffusion Web UI. To use them, you simply put model files inside {stable-diffusion-web-ui-path}/models/Stable-diffusion/ folder.

Reload models in Stable Diffusion Web UI, and you can use them in the model dropdown list.

Model extension names are like:

• .ckpt
• .safetensors

LoRA Model

LoRA: Low-Rank Adaptation of Large Language Models is a novel technique introduced by Microsoft researchers to deal with the problem of fine-tuning large-language models. [reference]

With Stable Diffusion LoRA model you can create images for similar to the ones in the training dataset.

• You must first install Additional Networks extension for Stable Diffusion Web UI. It is available in Extensions tab of Stable Diffusion Web UI.
• Then you simply put model files inside {stable-diffusion-web-ui-path}/models/Lora/ folder.
• After installing the extension, you should see an Additional Networks expandable control at bottom of the page. Open it and click enable button to use them.
• To use the actual model, just type <lora:filename:multiplier> in prompt input box. filename is the name of the model file without extension. multiplier is the multiplier of the model. It is optional. Default value is 1.

Embedding Model

Embedding is the result of textual inversion, a method to define new keywords in a model without modifying it. The method has gained attention because its capable of injecting new styles or objects to a model with as few as 3 -5 sample images. [reference]

To use them, you just put model files inside {stable-diffusion-web-ui-path}/embeddings/ folder.