Workflow name: AI e-commerce product model replacement

[Workflow introduction]
This workflow is divided into 3 parts. The first part uses Portrait Master and WD14 Tagger to reverse the prompt words. In Portrait Master, you can choose the nationality and modify the model's skin color. The second part controls the image through ControlNet, and then uses the SDXL model to generate the image. The third part uses the Flux model to optimize the image. Finally, the image is overlaid. At present, the built-in parameters of the sampler have been set, and you can run it directly. Come and experience it!

[Use scenario]
This workflow allows anyone to take the product as a model for shooting, realizing the regeneration of the model under the premise of ensuring that the product remains unchanged, eliminating the need for e-commerce products to find models for shooting and repair the image later.

[Key nodes]
instantID

[Model version]
SDXL
Model name: dreamshaperXl lightningDPMSDE.safetensors
FLUX
Model name: flux1-dev-fp8-Kjai.safetensors

[LoRA model]
None

[ControlNet application]
control-lora-depth-rank256
Start time: 0
End time: 1
Strength: 0.6

control-lora-canny-rank256
Start time: 0
End time: 1
Strength: 0.5

[K sampler]
CFG: 2
Sampling method: dpmpp sde
Scheduler: karras
Noise reduction: 1