Workflow name: Generate consistent models for custom poses of e-commerce clothing
[Workflow introduction]
Workflow operation process: First open the workflow, upload an image on the far left of the screen, and use the controlnet skeleton model to generate a unified character. Then split the 4 pose images and use SegmentAnythingUltra_V2 for matting and overlay processing. Then implement image style transfer through the flux1-redux-dev model, and then use the flux1-fill-dev model to redraw the mask area, and finally generate the effect map by the sampler. This workflow has many nodes and is divided into 2 parts. At this stage, it focuses on multi-pose model clothing migration and ensures consistency.
Workflow status and application suggestions: The current picture is a bit blurry. It is recommended to run the high-definition repair portrait workflow. The link can see the workflow note introduction.
[Usage scenario]
This workflow can accurately transfer the specified clothes to the generated model to ensure the consistency of the character, which is very suitable for the use needs of people in the e-commerce clothing field.
[Key nodes]
FLUX redux dev
[Model version]
FLUX
Model name: flux1-redux-dev
[LoRA model]
Model name: FLUX. 1-Turbo-Alpha
[ControlNet application]
Start time: 0
End time: 0.6
Intensity: 0.4
[K sampler]
CFG: 4.2
Sampling method: deis
Scheduler: beta
Noise reduction: 1
Sign In