Greg Rutkowski Style

Krea 2 version trigger word: grgw style

Recommended settings:

• 12 steps

• cfg 1.0

• euler / euler_a + beta57 ( er_sde + simple works too but less detailed)

NB: if you use Lonecat's Pro Grade Krea 2 workflow, use v6.

I recommend downloading the bf16 variant. Same quality, half the size.

Trained on Anima Base v1 using the example dataset and the anima_lora script with default settings. Trigger word: @greg rutkowski

Now that I've figured out how to use the training scripts, I realize v3 was way overcooked. Please use v4 instead!

v3 is deprecated

This is my first LoRA, and there are a few known issues:

• I have no idea what I'm doing, or why the angel's halo looks that. But two halos are better than one, no?

• Your gens might look underfit or overfit, depending on the seed.

• I have only tested this LoRA alone with Anima Base. If you use a fine-tune or other LoRAs, your mileage may vary.

If those are not deal breakers for you, go ahead and download this. Be sure to share your masterpieces and feedback :)

Usage tip: you can reduce errors by being highly specific with your prompts. In the Frieren example, I used from side to prevent the character's torso from turning the wrong way, and centered around the head of her staff to force the staff in the correct orientation.