Description
This is a massive ComfyUI workflow that helps to improve and simplify detailing and upscaling of stable diffusion images by enabling easy and automatic masking along with prompt refinement for multiple characters at once, allowing users to do highly selective inpainting/detailing/upscaling without need for any manual masking, and minimal prompting.
Fully updated for Krea2, it's now leaner, more efficient, and more powerful than ever, fully capable of repairing images using multiple character LORA, featuring point-and-click character selection, and the ability to completely transform subjects into different characters altogether.
Features:
- Automatic segmentation and masking for using point and click selection with natural language with SAM3.
- Supports detailing up to four characters simultaneously.
- Supports up to 3 LORAs per character/detailer pass combination.
- Universal LORA and Upscaler LORA.
- Centralized control of key functions across all characters and/or body parts: Sampler, Steps, CFG, Denoise Levels and more.
- Previews for every step of refinement, allowing for easy pinpointing if something goes wrong.
- Up to 6 refinement passes per character: full body, face and hair, clothing, breasts, genitals, and feet.
- Capable of re-rendering a subject as a completely different subject altogether. Extremely powerful in combination with character LORAs.
Uses
The workflow primarily enhances the fine details and overall quality of stable diffusion images. This update performs well at denoise levels of up to 60%, allowing it to do complete character replacement while still maintaining quality.
How to Get Started with the Model
Simply drag and drop the JSON file into ComfyUI, then follow the instructions after installing the extensions. It will download SAM3 for you.
Read the notes in the workflow for an explanation of what everything does.
The workflow takes about 3-5 minutes to run for four characters on an RTX4080, so you'll want to use it to improve base gens that you're already otherwise happy with.
Credits:
Big thanks to catlover1937. This workflow is a massive expansion and adaptation of the "Face Detailer" and "SD Upscale" sections of their excellent (and much more reasonable in scale) ZIB + ZIT Simple Workflow. This workflow would not have come into existence without that essential DNA to work from.
Description
The Ultimate Detailer Workflow is back for Krea 2, with a substantially redesigned and improved user experience, a leaner and faster implementation, and a vastly enhanced array of capabilities thanks to being powered by Krea 2.
- Now segmentation is fully powered by SAM3 - Yolo models are completely gone.
- Now supports hybrid prompting for segmentation. Click and drag on your image to select your character, then describe what parts of that character you want to segment. It's done in seconds. No more struggling with the vision model selecting the wrong person, or dealing with trying to figure out who corresponds to which index. You just tell it exactly what to do and it does it.
- Much better support for character loras.
- The ability to completely transform a character into another character, while still retaining features from the first character.
- Easy ability to repair images using multiple character LORAs to remove the 'bleeding' of attributes from one character to another
- Stable at up to 60% denoise levels.
- Faster and more efficient implementation. On an RTX4080, a full four-character detailing job with upscale takes around 3-5 minutes.
- Improved layout and organization to make it easy to get started>
Check out the example images, and give it a try! It's quite remarkable.