Three-mode ComfyUI workflow for Krea 2 with shared model, CLIP, VAE, and LoRA handling.
Includes:
Text-to-image generation
Reference-image generation
Face/head identity swap
The workflow uses a mode selector so only one pipeline runs at a time, helping avoid unnecessary VRAM use. It is designed around Krea 2 Turbo-style workflows and includes shared LoRA trigger handling, reference conditioning, identity editing, and configurable samplers.
Source images shown in Load Image nodes are not included with the workflow. Replace them with your own images.
Model files, LoRAs, and custom nodes are not included and must be installed separately.
Intended as a flexible starting point rather than a fixed preset. Sampler settings, resolutions, model choice, and LoRA combinations can be adjusted to suit your setup.
Very easy to patch in any form of upscaling you like on any of the pipelines.
I might release a more complex version at somepoint - or if you ask.
Description
Initial release of Atlas Krea2 Three-Mode Generator.
Includes Text-to-Image, Reference Image, and Face/Head Swap pipelines with shared model, CLIP, VAE, and LoRA handling.
Only one mode is intended to run at a time via the mode selector.
Source images and model files are not included. Replace image inputs with your own and install the required custom nodes/models separately.