This tutorial is collected at: Summary of AIGC tutorials from entry to mastery
If you have the basic knowledge of Stable Diffusion, making AI cyber locomotive maps will be smoother. Even without this foundation, as long as you strictly follow the steps of the tutorial, you can make it successfully.
The focus of this tutorial is to guide the operation process, without in-depth analysis of the functional principles of Stable Diffusion.
If you want to learn SD basics in detail, you can read this article: AI Painting: Stable Diffusion Ultimate Alchemy Book: From Beginner to Master
Effect picture preview:
complete process
Making an AI cybermap consists of the following 6 parts
① Model selection
②Prompt words
③Basic parameters
④ControlNet adjustment
⑤ High-definition processing of post-production pictures
⑥Cut and make a video
All the used models, prompt words and other materials can be found in the network disk link
Download link: https://pan.quark.cn/s/a8e797fe793b
detailed introduction
1. Model selection
The first comix v3.1 model. In addition, the three models of Baichengzhu Integrated Mecha, majicmixRealistic, and revAnimated are also good choices. In particular, it should be emphasized that there is no absolute best model selection, and each large model will appear differently. Therefore, you need to find the model that best suits your needs through trial and error and continuous exploration.
2. Prompt words
We provide some general prompt words (as shown in the picture).
If a more detailed description is required, other prompt words can be added according to the elements in the figure.
These prompt words have been uploaded to the network disk link. It should be noted that the prompt words I provided may not be the optimal choice, and we need everyone to explore and improve together.
3. Basic parameters
①Number of iteration steps: 30 (28-32 can be tried)
②Sampling method: DPM++ 2M Karras or DPM++ 2M SDE Karras
③The aspect ratio does not need to be set, it will be automatically adjusted later
④ The total number of batches is set to 4 batches, and the probability of drawing cards is higher
⑤Random number seed: -1
Four, ControlNet adjustment
① Turn on Enable, Pixel Perfect (perfect pixel), Allow Preview (allow preview)
②ControlNet model selection: Lineart or Tile. The sampling method of the former is line draft, and the sampling method of the latter is to blur first and then zoom in
The two styles will come out completely different styles. Generally, try both to see which one works better.
③Click the blue frame to replace the size of the picture after finishing, but pay attention to whether it will be too large, if it is too large, divide the length and width by 2
Finally, click Generate to output the picture. For the specific method, you can watch the following video tutorial to understand it better.
Five, high-definition picture processing
Find the [Additional Features/Post-processing] option
Zoom in and use [R-ESRGAN 4x+], click generate, you can change the picture to high-definition
Six, common problem solving
1. How to install Stable Diffusion
① You can read my article to learn, or you can search online search engines to find solutions
AI Painting: Stable Diffusion Ultimate Alchemy Book: From Beginner to Master
②If it is A card, mac system, low computer configuration, etc., you can consider using the cloud platform to operate SD
2. How to install the large model
The SD root directory here is the folder where you store SD when you download
3. How to install the aspect ratio plugin
①Find the [Extension] option
②Click to load the extended list
③Search 【Aspect】
④ Install【Aspect Ratio Helper】
4. How to install ControlNet
Controlnet data network disk preview:
Link: https://pan.quark.cn/s/47bc8c79892a
① Download the preprocessor
Download the "1. Preprocessor" folder in the network disk
Inside there is a "downloads" folder
Copy the entire folder to the SD folder
Specific location: SD file > extensions > sd-webui-controlnet > annotator
②Download model
Download the "2. Model" folder in the network disk
There are 14 models inside, copy these models to the SD folder
Specific location: SD file > models > ControlNet