Function
You can generate an image in reference to a particular posture using Stable diffusion and ControlNet.
Procedures
1. Use a simple character image without background for the best result.
2. Resize the character image into smaller size if you are not using a high spec PC.
3. Open Stable Diffusion “txt to Img” tab.
4. Select the checkpoint you want for the desired style.
5. Select exactly the same size of the uploaded image. For example, if you are uploading an image of 512*512, you have to select 512*512 for the generation.
6. Add Control Net Unit 0 “Openpose” & add the reference image to the column.
7. Preview the posture mapping by clicking on “Run preprocessor” as below:
8. To finetune the posture mapping, click on “Edit” and change the coordination to the desired pose.
9. “Send posture to ControlNet” when it is done.
10. Once you have the posture prepared, Add Control Net Unit 1 “Canny”
11. Add the original character image to the column and input the following settings:
12. Write your ideal Prompt and Negative prompt, and click generate.
13. If the generated image is unsatisfactory, try to adjust the following factor:
– The prompt (be more specific and detailed and use brackets to emphasize)