Loading
Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
Trained on a dataset of 55 handpicked images.
Model source
Source description
Trained on a dataset of 55 handpicked images.
As always, I recommend euler + beta57
Krea 2 Version
Trigger word: sgca style
The bf16 variant produces the same quality output as fp32. However, there are minor differences in details. You have to play Where's Waldo to spot them, though. That being said, I personally prefer the fp32 variant after testing with side-by-side comparisons.
The bottom line: choose bf16 if you prefer smaller files. Choose fp32 if you've got the VRAM to spare.
Anima Version
Trigger word: @sgca style
Sources
1 sourceModel artifacts
2 artifactsSource excerpts
1 excerptSource context: 396 downloads · Type LORA · Base model Krea 2
Again, euler_ancestral fried all of my gens. I have no idea why. A quirk of T-LoRA, maybe?
