๐ข Introduction to Stable Diffusion Models
The model is the most crucial factor determining the quality of the generated images. Thus, learning how to utilize various models to enhance the quality and style of generated images is a valuable pursuit.
You can find various models here:
https://huggingface.co/models?pipeline_tag=text-to-image
Large Modelsโ
Also known as the base model, this category includes models with the largest training datasets, which significantly impact the quality of image generation. The default large model for Stable Diffusion is typically SD x.x (version number).
The downloaded large model can be placed in the application's models/Stable_diffusion path for use.
Anime Style Large Modelsโ
abyssorangemix3AOM3_aom3a1
dalcefoPainting_v4
pastelMixStylizedAnime_pastelMixPrunedFP16
Oil Painting Style
threeDelicacyWonton_threeflavorwontonmixv
Ink Painting Style
revAnimated_v11
Large Scenes
Counterfeit
TMND-Mix
Exceptional Indoor Environment
animelike2D
Anime Series Style
cetusMix
Color Box Model
Night Sky YOZORA Style Model
VAE Beautification Modelโ
Similar to applying a filter, choosing VAE is akin to adding a filter to the image; default is none.
Lora Modelโ
The Lora model is a smaller model generated from parts of the large model. It doesn't have the full capabilities of the large models but performs better under specific conditions after targeted training.
For example, if we see some Lora model images we like on C station, we can download its Lora model by clicking on the exclamation mark to view its parameters.
Once downloaded, we place the Lora model in this path
In the application, select the Lora model under additional network to use.
Below are the comparison images generated with and without using the Lora model.
