Central Room Group: Start here
The workflow operates within a central room group, reducing the need for travel between different locations.
This centralized setup enhances efficiency.
Lora Integration:
Lora are loaded into the model, 3 Lora loaders provided.
Trigger words associated with Lora are automatically fetched from CiviAi & added to the prompt.
Model Configuration and FreeU V2 Implementation:
FreeU V2, a sophisticated image generation technique, is employed.
The modelโs architecture and hyperparameters are optimized for high-quality output. (using Auto cfg for faster uncond)
Image Processing and Resemblance Enhancement:
The base image can be sourced from either a URL or a local file.
An IPA (Image Preprocessing Adapter) adapts the input image to the modelโs requirements.
Control nets are used to achieve a close resemblance to the desired output image. (tile based)
The system ensures that the generated image aligns with the intended visual style.
Latent Space Manipulation with Noise Injection:
Noise vectors are injected into this latent during image generation.
This introduces additional details and variations, making the output more interesting and diverse.
Image Storage and Naming:
The generated image is stored in a specified location (e.g., a designated folder ).
The filename is automatically generated using the WD14 Tagger, ensuring consistency and organization.
Optional Detailer:
Eye, mouth, and hand detailers can be selectively applied.
These detailers enhance specific features in the generated image.
Users have the flexibility to enable or bypass them based on their preferences.
Super-Resolution (SD Upscale):
Optionally, the image can undergo super-resolution scaling (SD upscale) to enhance its resolution.
This process increases the image size while preserving quality by 3x.
HDR Effect and Finalization:
As a finishing touch, an HDR (High Dynamic Range) effect can be applied to the image.
This enhances contrast, color vibrancy, and overall visual appeal.