This is an advanced workflow for Krea 2 Raw that uses the split sigma function to send a few steps of the raw model for the initial render and then finishes it with a turbo lora. There is a custom node that seems to do this but this is just a workflow using the standard custom samplers. This has a lot of flexibility to tailor the steps and crossover.
This also includes two detailers and image input nodes I have used on my previous workflow. There are a bunch of notes in the workflow itself with a lot of custom workflow in the subgraphs. A bit rough and ready so really focused on an experienced Comfyui user.
NOTE: This is using GGUF loaders so it is low ram focused out of the box. Replace them with standard loaders for larger models.
Enjoy!
Additional Resources:
https://huggingface.co/Comfy-Org/sam3.1
https://huggingface.co/vantagewithai/Krea-2-Raw-GGUF
https://huggingface.co/lvladikov/Krea2-Turbo-Distill-4step-LoRA
https://huggingface.co/lvladikov/Krea2-Turbo-Distill-2step-LoRA
https://huggingface.co/mradermacher/Qwen2.5-VL-7B-Instruct-abliterated-GGUF
https://huggingface.co/mradermacher/Qwen3-VL-4b-Heretic-GGUF
Description
Initial version