Workflow Name: Multi-Head View Role Table
【Workflow Introduction】
When starting the workflow, two images need to be uploaded. One is a character outfit image, which should clearly showcase the character's clothing and styling details, including garment design, color combinations, etc. The other is a skeleton image (can use the built-in one). Based on Portrait Master and Runninghub Captioner, reverse-engineer prompt combinations to obtain forward prompts for us. Use control net to control posture through the skeleton. Combine the FLUX model and FLUX LoRA model for image generation.

【Use Case】
The FLUX model will generate 4 images with different postures based on one image, while maintaining relatively similar facial features. This undoubtedly brings you great surprise and experience.

【Key Nodes】
Portrait Master and Runninghub Captioner

【Model Version】
FLUX
Model Name: fux1 dev fp8 Kijai.safetensors

【LoRA Model】
Model Name: 盛世美颜系列.safetensors

【ControlNet Application】
FLUX.1 dev ControINet Union Pro Shakker Labs.safetensors
DW Posture Preprocessor
Start Time: 0
End Time: 0.4
Intensity: 0.6

【Custom Sampler (Advanced)】
Sampling Method: deis
Scheduler: beta
Denoising: 1