Workflow name: [Consistency] Generate character material set
[Workflow introduction]
This workflow is very large. It costs 200 coins to run it all on RH. If you only want to train the head, please turn off the half-body and full-body workflows.
You need to upload 2 images to open, one is the portrait and the other is the clothes image. See this workflow from the far left.
Step 1: Perform facial migration through Apply PuLID Flux;
Step 2: Infer the character image and clothes image prompts through Runninghub Captioner, and then write some prompts by hand;
Step 3: Control the skeleton through Controlnet. Here I have built-in 3 images, namely head portrait, half-length portrait and full-length portrait;
Step 4: Process the generated image through SD amplification to make the picture clearer and more detailed;
Step 5: Segment the amplified image through this node;
Step 6: Because of the material of the LoRA training set, I have overlaid the image again so that the final image is in the positive direction;
The last step: Choose an angle you like and use the Expression Editor node to make facial expressions to make this material set look richer.
If you still want to process the image with lighting effects, it is recommended to change to another flow: IC-Light custom lighting flow, the URL is: https://www.runninghub.ai/post/1857376758420156417
[Use scenario]
If you want to generate a LoRA training set from 1 image, you can try this flow.
[Key nodes] PuLID Flux, Controlnet
Model version]
FLUX
Model name: fux1-dev-Q8_0.safetensors
[LoRA model]
None
[ControlNet application]
FLUX.1-dev-ControINet-Union-Pro-Shakker-Labs.safetensors
DW attitude preprocessor
Start time: 0
End time: 0.85
Intensity: 0.85
[Custom sampler (advanced)]
Sampling method: deis
Scheduler: beta
Noise reduction: 1
Sign In