Deployment Package Author: Akiba aaaki
Disclaimer:
This installation package and launcher are provided for free without any profit purpose
Hello everyone, I am rain or shine.
As we all know, StableDiffusion is a very powerful AI drawing tool. Friends who need to know more about StableDiffusion can check out my previous article: Midjourney VS StableDiffusion, two of the most popular AI drawing tools recently .
What I bring to you today is the introduction and installation of Stable Diffusion Research (2) sd model ControlNet1.1.
First of all, let's clarify what exactly is this ControlNet1.1?
ControlNet1.1 is a Stable Diffusion painting plug-in. It updated the V1.1 version in April 2023, released 14 optimized models, added multiple preprocessors, and made some optimizations on the original models, so that It's more functional than before.
So let's take a closer look at the new model and new functions of ControlNet1.1.
1. The new model and new function of ControlNet1.1
1. Add Lineart model and lineart_anime model
Now two new models have been added, lineart and lineart anime, which can completely replace the original Canny's powerful line draft coloring. Both black and white and color coloring functions are very powerful.
2. Add Instruct Pix2Pix model
The model can accept an image and corresponding text instructions (that is, prompt), edit the image according to the instruction, and have the ability to partially manipulate the image as a whole.
Change day to night, as shown in the figure:
Change the hairstyle, as shown in the picture:
Dress up, as shown in the picture:
3. Add Tile model
The Tile model has a good effect on high-definition restoration of pictures, enhancement of details, and automatic inference of content based on the picture. According to the official introduction, there are 5 functions:
- Image upscaling 2x, 4x or 8x super resolution possible
- Image detail can be added, changed or regenerated in the image
- Possibility to repair, refine and improve poor image details obtained by any other super-resolution method
- It can boot SD zoom script, generate wrong problem
- Unfinished artwork can be completed if these are drawn with color blocks
Before image restoration:
After image restoration:
Before detail adjustment:
After detail adjustment:
4. New shuffle model
The shuffle model can be called a reorganization model. From the perspective of the use effect of the shuffle model, the shuffle model can generate new pictures by obtaining the style of the original picture and using this style, so when we find a better reference picture, we can Use the shuffle model for image creation.
Let's take a look at the effect of character map processing:
From the perspective of the image forming effect, the newly generated image style, color tone, and some poses of the characters are restored and preserved. Therefore, we can use the Shuffle model to use the original image for reference and reorganize the image to generate a new image.
Let's take a look at the processing effect of the landscape class image:
Judging from the finished picture, the main color scheme and style refer to the style of the original picture, and there are also some references to the composition structure.
This is the end of the introduction of new functions and model effects. Next, I will introduce how to install and use ControlNet1.1.
Before, we have been using the Control plug-in. In fact, we only need to upgrade the plug-in to the new version to use the new version of the model. So we just need to update the plugin, then download the new model and put it in.
Two, install ControlNet1.1
There are two main points in this upgrade and installation:
-
Upgrading the ControlNet plug-in version
-
Models with various preprocessors installed
If you have not installed the Stable Diffusion integration package, you can read Stable Diffusion Research (1) sd integration package v4.2 version installation instructions This article, learn about the computer configuration required to install the Stable Diffusion integration package and download the relevant installation package.
First of all, you need to prepare the ControlNet 1.1 installation file. The file is integrated in the integrated package sd-webui-aki-v4.2 version . If you have not downloaded it, you can download it yourself.
1. Upgrade the ControlNet plug-in version
The first step is to enter the integration package directory: sd-webui-aki-v4, double-click the A launcher.exe directly to run
After startup as follows:
The second step, click on the version management on the left
The third step is to switch to the extension tab and click the "One-click update" button in the upper right corner.
2. Install various preprocessor models (optional)
The first step is to copy the download folder in the preprocessor model folder to the annotator folder under the plug-in directory.
\controlnet1.1\预处理器\download
\sd-webui-aki-v4.2\extensions\sd-webui-controlnet\annotator
As shown in the picture:
The second step is to put the controlnet1.1 model into the sd-webui-controlnet\model folder.
\controlnet1.1\模型
\sd-webui-aki-v4.2\extensions\sd-webui-controlnet\models
After completing the above operations, close the launcher, console, and webui, and then restart to take effect.
As shown in the picture:
3. ControlNet1.1 user experience and some changes relative to the boss
1. The color inversion mode has become a separate preprocessor
2. Add perfect pixel mode
Selecting Later can resolve some resolution issues with ControlNet processing.
3. New control mode
New control mode:
- balanced
- more prompt
- Prefer ControlNet
When choosing to be more biased towards ControlNet, you need to reduce CFG.
4. The HED of the boss has changed to SoftEdge
5. Added several preprocessors
OpenPose preprocessor -> OpenPose model
Openpose_full, one model gesture, hand and face are all done
6. The Depth model of the boss -> the new version recommends Leres++, Zoe preprocessor
Well, that's all for today's sharing. More functions and changes of ControlNet1.1 will not be described in detail here. Interested friends can study by themselves. In the next article, I will bring you detailed usage methods of each functional module of sd, so stay tuned.