v 2.0
This is the 8-Step Turbo-LoRA for Anima-Preview2. Also works with Anima-Preview3.
Recommended Settings:
Steps: 8
CFG Scale: 1
Samplers: er_sde or res_2m, res_multistep
Original workflows are included in the examples. Please experiment and share your feedback. Training was conducted using 100% renewable energy.
v 1.0
This is the 14-Step Turbo-LoRA for Anima-Preview2. Still a bit experimental, this fun version demonstrates turbo-training capabilities within the Anima architecture. It was distilled on 227 Anima-Preview-2 latents at 2000 steps.
Recommended Settings:
Steps: 14
CFG Scale: 1
Samplers: er_sde or res_2m
Original workflows are included in the examples. Please experiment and share your feedback.
Just remove unnecessary disabled nodes in the examples.
Description
This is the 8-Step Turbo-LoRA for Anima-Preview2
Improved Style Diversity
Baked Negative Prompt: worst quality, low quality, score_1, score_2, blurry, jpeg artifacts, sepia, defective
Dataset: 337 distilled, cherry-picked latent images with high diversity.