UPDATE: V2 features base model training (instead of preview3 on V1) while still merging with the base model. Revised V1's dataset removing a few dozen images, replacing and adding more high quality images. Changed the entire captions/tags for the entire dataset.
HighBlend is a twist on OverBlend! HighBlend focuses more on high quality, realistic Blender images instead of the more smooth and Overwatch feel of OverBlend!
Uses Anima Base.
YOU NEED:
qwen base encoder & qwen image vae
SETUP:
Steps: 30 - 50
CGF: 4 - 5
Sampler: Eular/Eular A/Res 2s - Simple/Beta/Karras
Gallery images have included metadata workflow (V2). Mp4 uses checkpoint as starting images.
Description
Revised dataset. Replaced a few dozen and added on top of the existing dataset. Now trained with the Base model & merging with the base model. V1 used preview 3 as the training model then merged it the base model. Changed all tags/captions for entire dataset.