💫Mixer
Using Mixer, you can replace the background of your input image while retaining the main subject and generate corresponding background images, replace the face in a photo, or mix any 2 images together.
Replace Background
Start by uploading an image that you'd like to change the background of.
Prompt: Describe the prompts for the background you want to generate, such as:
Negative Prompt: Specify the prompts for content you don't want in the generated image.
Classifier-Free Guidance (Scale): This parameter determines the strength of the prompts, indicating how closely the AI should adhere to the provided description. A higher value makes the generated image closer to the prompts.
Product Size: Control the size of the resulting generated image. Here, we'll use "Original," which means the generated image will have the same dimensions as the original image.
Number of Inference Steps: 30
Seed: 9589360
Finally, you will get the image below. As you can see, the main subject of the image is retained while the background has been replaced with the desired one:
Face(s) Swap
For photos or images with just one face, simply upload the source image along with a clear, front-facing photo of the face. The resulting image will appear as follows:
For photos or images containing multiple faces, you need to upload the source image and then upload a clear, front-facing photo of the face (face 1), choose the first face you want to swap; and then upload another clear, front-facing photo of the face (face 2) and choose the second face you want to swap. You can swap up to two faces in a single image.
Mixing Character and Background Images
You can mix one image contains the character and the other image contains the background following steps below:
Prompt:
Negative Prompt:
(Denoising) Strength: 0.75
Classifier-Free Guidance (Scale): 7.5
Scale: 2 (Proportional to the size of the character image in the resulting image)
manual_seed(random): Using the same seed with the same options like prompts, guidance scale, etc., will consistently produce the same image. This is useful for fine-tuning previous creations, understanding how other options affect the process (by removing the "random" element), or sharing your work.
The mixes image is as follows:
Mixing 2 Images
Or you can mix 2 separate images from left to right following steps below:
Prompt: Your prompt for the mixed image ouput:
Prompt_L: Your prompt for the left image:
Prompt_R: Your prompt for the right image:
(Denoising) Strength: 0.75
Classifier-Free Guidance (Scale): 7.5
Controlnet_Scale: 1.3
Number of Inference Steps: 30
Seed: The seed for generating the image. Using the same seed will result in images with similar styles.
Randomize: Enable this switch to ensure that the seed for generating images is different each time.
Sampler: Sampler
K_Eular | If you just want to get some simpler results |
---|---|
DDIM | Focuses on speed, fusion, novelty, and good quality results, Step Range: 20-30 |
DPMSolverMultistep | Expects high-quality images and doesn't care if the images converge or not. |
Control_type: Control the position of the fusion of two images
Controlnet: Choose the controlnet to use
canny: Takes the line drawing of the input image and outputs a new image
depth: Computes spatial position information for the image to generate images with the same depth of field
hed:
The final mixed image outpt is:
Last updated