I. Introduction
Recently, AI image generation has attracted attention. It can generate beautiful images based on text descriptions, which has greatly changed the way people create images. As a high-performance model, Stable Diffusion generates higher image quality, runs faster, consumes less resources and occupies less memory, and is a milestone in the field of AI image generation.
However, it is also a troublesome thing to deploy and experience Stable Diffusion, either because of network limitations or because the GPU configuration requirements are too high. For users who do not have a high-end GPU, they can only go to the big platform for free. They have deployed it with Google Colab, but the steps are cumbersome and they are easy to drop.
Recently, I came into contact with Baidu's Flying Plasma platform, and configured it to run and experience Stable Diffusion for more than a week. It is easy to get started, stable and does not drop the line. The key is that it can run for free for 8-16 hours a day!
Two. Text
1. Register and log in
If you are already a user of Feijiang platform, please skip to the second step directly. New users go to the official website of Feijiang to register and log in, and then complete the personal information.
2. fork project
There are several Stable Diffusion projects on the Flying Slurry platform, and I rewrite it based on one of them, which simplifies the process steps, and comes with 5 large models and 9 Lora styles, click to enter the project .
Click 'Run' , wait for the copy to complete, then select 'Start Environment' , select the environment with GPU , you can choose 16GB-16 hours, or 32GB-8 hours, you can choose according to your needs, confirm, wait for the startup to complete
After the startup is complete, you can enter the console environment.
3. Running experience
You can see the following interface, which contains instructions and code blocks, and can be completed by following the steps inside.
Execute the code and select the play button on the left of the code.
Follow the steps, rely on installation - model decompression - restart the kernel - running experience, you can see the UI interface indicating success.
At this point we can enter a positive description (although it supports Chinese, it is recommended to translate it into English for better results), select the size and other parameters. Click the 'generate picture' button below to generate it, and the generated picture will not be displayed too much, please 'do your own thing, keep yourself clean' !
So far, the tutorial is over. If you have any questions, you can leave a message in the project or comment area. If you think the project is good, please click a little red heart for the project!