This Turbo LoRA is extracted by taking the delta difference of Anima Turbo v1.0 and Anima Aesthetic v1.0, then compressed to a 128-rank LoRA.
This works well on Anima Aesthetic-based checkpoint, and half-well on Anima Base-based checkpoint.
Start with CFG 1 and 8-12 steps
Description
Init and probably the only release
FAQ
Comments (2)
Thanks. But isn't file size little to big compared to other turbo loras?
It's because I'm trying to keep ~98% of the original quality from the full rank (~3.18 GB).
I can drop it to rank 64 or even 32, but from my manual check, the quality is very lossy.
LMK if you think you or other people will still interested in rank 64 or rank 32 version.


