Stable Diffusion (SD), a popular deep learning text-to-image conversion model, allows you to create detailed images based on text prompts. But how impressive and detailed your images turn out depends on how specific your text prompts are.
Developing great prompts involves a lot of experimentation. In this article, we will discuss some settings to dramatically change image variations and how to configure SD to run locally on your PC.
How to create a great message for stable diffusion
“Quick brewing” requires time and experimentation to achieve the best results. You should be as specific as possible and clearly define your artistic styles or mediums and particular artists. Also, avoid keyword interference.
Next, it’s important to know how to customize SD feature settings to automate your testing.
The advice that follows is based on a local SD installation; however, they are also applicable if you run the online version. For detailed steps on how to install SD locally, please scroll down to “How to Set Up Stable Broadcasting on a Windows PC” section.
Make sure your prompts work
Before you begin, make sure the prompts are ready by doing the following:
- In the “txt2ing” tab, leave the message field empty and scroll down to the “Script” section.
- Click the drop-down menu and select “Ask from file or text box.”
- You can place the text file of your instructions in the “File with entries” window. Alternatively, check the “Show text box” option and enter your directions in the “Messages” window. A text file is probably easier to work with, as it is easier to modify and save.
- Go to the location of your text file and drag it into the window. If you make changes to the file, you will need to place the updated file in the prompt window, as the user interface will not update it automatically.
- In the “Seed” input field, set your custom seed and then click “Generate”.
If you’re willing to experiment using art generated by your prompts, we’ll start by using the graderless guide scaling (CFG) function.
Experimenting with the CFG scale
The CFG scale measures how closely you want the model to adhere to your message when generating related images. For example, a CFG scale value of “0” will generate a random image based on the seed. On the other hand, a CFG scale of “20” and a maximum SD will create the closest possible match to your message.
Follow these steps to experiment with your message using the CFG scaling feature:
- Go to “Script” and then choose “X/Y Chart”.
- Click the “Type X” drop-down menu and select “CFG Scale.”
- From the “Y Type” drop-down menu, choose “Steps.”
- In the “X Values” input field, consider setting the CFG scale to “3-5”. This will generate integer versions of your image. If you want to produce half numbers, enter: “3-5 (+ 0.5) using square brackets.
- Then use the “Steps” in the “Y Values” field to define the number of variations between the range. For example, to try 10 to 40 steps, enter “10-40.” To use five variations, enter “10-40 [5] with brackets.
- For a clear overview, make sure “Draw Legend” is checked.
- Click the “Generate” button.
Depending on the variations you order, you will receive many rendering options. Also, please note that all images are in full resolution, so the version you like will be the finished product.
Experiment with the quick matrix
The “Prompt Matrix” is another powerful way to test your prompts as you generate more variations from the same prompt. Follow these steps to customize the message array feature:
- Go to the “Script” drop-down menu and choose “Message Array.”
- Enter the message in the message field, then press the space bar. Enter a pipe character: “|” – then add another space. Enter the different style versions you want to use, for example, “oil painting” or “watercolor,” and use the vertical tube to separate each one.
- Once you press “Generate”, depending on the number of variables entered, that number multiplied by itself will be the number of variations that will be displayed. For example, 4 arguments X 4 = 16 results.
Experiment with sampling methods
The sampling method refines your image from noise to recognizable shapes. Follow these steps to test sampling methods:
- Go to the “Y Type” drop-down menu and select “Sampler.”
- In the “Y Values” text field, enter the sample method, for example, “Euler a,” followed by a comma to separate other sample names. Consider testing in at least three ways.
- To keep things simple, set the “X value” of the CFG scale to three variations by entering “3-5.”
- Press the “Generate” button.
How to Set Up Stable Broadcasting on a Windows PC
If you have 15-20 GB of free disk space, you can access the SD for free by installing it on your PC. The following steps include direct access to all the files you need to install and download.
To make the process easier, consider creating two folders, one to save all downloaded SD files and the other to install your local version of SD. For example, you can create a folder under “Documents” and name it something like “SDLocal” without spaces, since SD can have problems with spaces in the directory name.
Download stable broadcast files
- First visit python.org to download the latest version of Python.
- Scroll down to the “Files” section, then select the version of the Windows installer (64-bit) to download it.
- Visit the git-local-branching-cheap “Downloads for Windows” page.
- Click on the “64-bit setup for Windows” option to download it.
- Now visit GitHub to download the Stable diffusion web user interface. Click the “Code” drop-down menu, then select “Download ZIP.”
- Go to the huggingface.co website to download the latest SD version. On this website, you will need to create a free account. Scroll down to choose the “…EMA Full…” version for a full version. This is a large file and may take a while to finish downloading.
- Visit the “GFPGAN” GitHub website to download the Gen files for the “Facial Restoration” feature. Scroll down to the “Updates” section with green boxes checked. Download “model V1.3” to get more natural results; Also, please download “model V1.4” for more details.
- You can also download “Notepad++” from the Notepad++ website. Click the latest 64-bit version at the top.
Install stable broadcast files
- First, go to your Python installation file, double-click it, and then select “Install now” in the pop-up window. Click “Close” once the installation is complete.
- Go to the “Stable Diffusion Web UI Master” zip file, double-click it, and then double-click the version folder. Select all the files inside that folder, then move them to the local SD folder you created earlier in the “Documents” folder. Make sure all files are visible there.
- Return to your folder containing all the downloaded files. Find the “GFPGANv1.3.pth” and “GFPGANv1.4.pth” files, then drag them to join the other SD files in your local SD folder.
- Find the file “sd-v1-4-full-ema.ckpt” in the downloaded files folder, and then click the file name to change it to “model”. Drag it to your SD folder.
- Double-click the Git .exe file, accept the license, then continue clicking “Next” until you reach the “Installation” screen, then click “Finish” once you’re ready.
- Locate the notepad++ installer .exe file to install it, and then click “Finish.”
- From the local folder of your SD, click on the “webui-user.bat” file and a pop-up window will appear saying “Windows protected your PC”. Click “More Information” and then “Run Anyway.”
- The command prompt window will open and display the error “Python could not be started.” Leave the command prompt window open.
To clear this error, you will basically need to connect Python to SD; that is how:
- Right click on the “webui-user.bat” file, go to “Show more options” and then select “Edit with Notepad++”.
- Notepad++ will open. Click the search icon and enter “Phyton.” The Python application will display the results.
- Right click on the Python application and select “Open file location”.
- The Python folder will be displayed; Right click on the Python 64-bit shortcut and then select “Open file location”.
- Right click on the “phyton.exe” file, then select “Copy as path”. The route will be copied to your clipboard.
- Go back to Notepad++ and place the cursor next to the “set PHYTON=” line to copy the path. Click “Save.”
- Go back to the local folder of your SD, then double click on the “webui-user.bat” file.
- This process may take a while and will not show you any indication of progress. Once the installation is complete, the Command Prompt window will be filled with information from the SD. This window should remain open when running SD locally.
- The most important information is the IP number located in the “Running at local URL:” line toward the bottom of the screen.
- Copy the URL and then paste it into your browser’s address bar to access the local version of SD.
The art of creating a great message SD
When it comes to working with AI art models in SD, the best images are created from specific, well-written prompts. There are even great sites that allow you generate images for stable online dissemination. But quick crafting requires time and a lot of experimentation with features to see how SD responds to particular settings. The more time you spend playing with SD, the more you realize that getting it to generate the art you want is an art in itself.
Have you managed to create images with SD that impressed you? Tell us about some of your favorite works of art in the comments section below.
