Models, sometimes called checkpoint files, refer to pre-trained image generation algorithms that are used to generate general or subject-specific images.

Which images the models can generate depends on the data used to train them. If there were never cats in the training data, the model would fail to generate images of cats. Likewise, if you train the model with only cat images, it will only generate cats.

This article will take you through some common models and how to install, use and combine them.

• Common models (v1.5, Anything V3, Open Journey v4)
• Recommendations for some specific style models (DreamShaper, ChilloutMix, Robo Diffusion, Mo-di-diffusion, Inkpunk Diffusion)
• installation and use
• Merge models and merge examples

common model

Stable diffusion v1.5

v1.5 was released by Runway ML in January 2022. The model was further trained on v1.2. It produces slightly different results compared to v1.4, but it's unclear if they are better. Most of the time, unless you're really picky about certain styles, using it as-is will suffice.

Anything V3

Anything V3 is a special-purpose model trained to generate high-quality anime-style images.

You can use danbooru tags (such as 1girl, white hair) in the prompt.

Through instructions, he can generate anime images with a unique style very well.

But one downside (for me at least) is that it produces women with disproportionate body shapes, so it may need to be corrected with other models.

Open Journey

Open Journey is a very useful general model, he made some fine-tuning on the basis of Mid Journey v4.

Special model recommendation

Dreamshaper

Dreamshaper is fine-tuned for portrait illustration styles that fall somewhere between photorealistic and computer graphics.

Chillout Mix

ChilloutMix is a special model for generating photo-quality Asian women.

It may produce some nude images, so you need to use words like "nude" in the negative prompt.

Robo Diffusion

Robo Diffusion is a fun robot-inspired mod that will turn every one of your subjects into a robot!

Mo-di

If you want to generate some Pixar-like style, this mockup is for you.

Inkpunk Diffusion

Inkpunk Diffusion is a Dreambooth trained model with a very unique illustration style.

How to install and use the model

To install the model in AUTOMATIC1111 GUI, take the model (.ckpt) file and place it in the following folder

stable-diffusion-webui/models/Stable-diffusion/

Click the "Reload" button to the right of the model selection

You should see the checkpoint file you just dropped in available for selection.

Alternatively, you can press the "iPod" button under "Generate".

The corresponding model will also appear.

Merge the two models

As shown in the figure, these models can be mixed on the Checkpoint merger page

Multiplier (M) is used to adjust the relative weights of the two models. Setting it to 0.5 will merge two models with equal importance. After pressing "RUN", the new merged model will be ready for use.

Here is an example image combining F222 and Any V5 with the same weight (3.0):

Proficient in the unique styles of the weekly models, and good at mixing them, you can create a lot of "blue is better than blue" pictures.

Have a good time~