1. Introduction
2. Computer Requirements
3. Installation Guide
4. Model Installation
1. Introduction
Currently, there are two authoritative AI drawing software available on the market that can be used for work. One is called Midjourney (hereafter referred to as MJ), and the other is called stable-diffusion (hereafter referred to as SD).
MJ is not free and requires a subscription of $10 per month. For ordinary people who just want to try AI drawing, it’s not cost-effective. On the other hand, SD is open-source and free, but it has a high learning curve and installation can be a challenge for many. It also requires a powerful computer configuration (graphics card, memory).
If you can solve the installation issues, learning challenges, and computer performance, you will gradually discover that the playability of SD is incredibly high, with the most direct advantage being: it doesn’t cost money and is not restricted by the internet 😆.
Due to its open-source nature, SD has many free high-quality external pre-trained models (fine-tune) and plugins, such as Controlnet, which can extract object outlines, human pose skeletons, image depth information, and perform semantic segmentation. Using it allows us to precisely control details like character movements, gestures, and composition during the drawing process.
The Mov2Mov plugin can stylize real videos; SD also has Inpainting and Outpainting capabilities, allowing for intelligent local modifications and extensions of images, which are currently not possible with Midjourney.
For example, Twitter user @Toyxyz3 achieved precise control over finger and pose in SD by rendering hand mesh depth and open pose skeleton separately.
You can also use SD LORA or Dreambooth to train models of your favorite characters or art styles, creating your own exclusive AI drawing tool. Some artists and designers train models with their original works, and when inspiration strikes, they can quickly generate various concept sketches using text descriptions with SD.
https://github.com/Stability-AI/stablediffusion
2. Computer Requirements
The core points of computer configuration are: check the graphics card, memory, hard drive, and CPU. The most important is the graphics card.
NVIDIA graphics card (AMD cards are not supported), with a minimum of 10 series, at least 4GB of VRAM, 6GB is acceptable, and no upper limit; minimum 8GB of RAM, 16GB is acceptable, and no upper limit; the hard drive should ideally have at least 500GB of usable space, with SSD being best, although mechanical drives are also acceptable. The CPU doesn’t have strict requirements; if you have a good graphics card, the CPU is generally not very poor.
How to check your computer’s graphics card?
Open Task Manager. Performance – GPU, and you can see the VRAM. As long as it’s greater than 4GB, you can give it a try. There’s a driver date on the right; any driver version and date from the past year is fine, just not too outdated, and updating to the latest is better.
How to check your computer’s memory?
Still in Task Manager, click on Memory, and you can see the memory capacity in the upper right corner.
3. Installation Guide
If you follow the open-source address above for installation, it may take one or two days for someone who has never touched code to get it done, and you might not even finish in that time.
If I were to write a beginner-level step-by-step guide, I feel it would need several articles: Python installation, Git installation, graphics card driver updates, and CUDA, webui………😂😂 it might just scare people away.
Fortunately, there are experts who have created one-click startup programs, such as Qiuye in China and A1111 abroad, which greatly lower the installation threshold. In this basic tutorial, we will first use the one-click startup package to quickly experience the fun of SD. The one-click startup package is just a visual one-click interface and does not affect the output quality; it only removes the local deployment barrier.
https://pan.baidu.com/s/1qWggi0qMMTdMpqCNCRcjoQ?pwd=a241
https://pan.baidu.com/s/1_oY7lqRqWn330yEjIIhTxg?pwd=6666
https://support.qq.com/product/488228
Download and extract
Click the launcher, and then click one-click start
On the first startup, a txt file will pop up. According to the requirements, copy and paste ‘I have read and agree to the user agreement’, save the txt
Then restart the one-click start
Startup successful, it’s that simple 😂😂
05
Try it out a bit; you can input this command and then click generate, and the image will appear in the lower right corner to see the effect.
oil painting with heavy impasto of a pirate ship and its captain, cosmic horror painting, elegant intricate artstation concept art by craig mullins detailed
4. Model Installation
By default, there is only one model available, which can be seen on the launcher model management page. The local checkbox must be checked to use it, while other models need to be downloaded to use. We can also see in webui that only this one model is available for selection.
How to import? When downloading the installation package, there is an optional-animefull-latest model included, using this as an example to demonstrate how to import models.
Click the add model button in the upper right corner, select the model with the .ckpt extension in the pop-up file manager, and click open
Successfully imported
Close the console and restart; webui will show the newly added model
The rest is for you to experience.
