Controlnet Union
☁️BizyAir Controlnet Union SDXL 1.0¶
☁️BizyAir Controlnet Union SDXL 1.0 integrates the project controlnet-union-sdxl-1.0 . With this node, you can generate images using various control types.
In conjunction with the controlnet preprocessors already provided by BizyAir, you do not need to prepare models and environments; you can directly perform tasks like image-to-image generation.
Here is an example: converting a photo of a great white shark into a depth map and then drawing a cyberpunk-style submarine. You can find similar examples in the repository under the examples section.
Note
The complete workflow can be found at: github.com/siliconflow//BizyAir/blob/master/examples.
Click the provided link above to get it.
Looking forward to sharing more of your creative uses, enjoy!
Parameter Explanation¶
Here are 6 types of inputs, each corresponding to a specific type of ControlNet control image, and they are all of the IMAGE type.
- openpose_image
- depth_image
- hed_pidi_scribble_ted_image
- canny_lineart_anime_lineart_mlsd_image
- normal_image
- segment_image
If you are unsure how to obtain such images, you can use the BizyAir ControlNet Preprocessor to generate them. You can use one or more control images simultaneously, for example, connecting both openpose_image and depth_image. However, from a practical view, it is not recommended to connect more than two control images simultaneously.
prompt and negative_prompt correspond to the positive and negative prompts, respectively.
- A higher value for
guidance_scaleresults in images that more closely follow the input prompt.
control_guidance_start and control_guidance_end are values between 0 and 1. They indicate the proportion of the total steps during which ControlNet is applied. For example, if the total number of steps is set to 20, and control_guidance_start=0, control_guidance_end=0.5, only the first 10 steps will be influenced by the control image.