What to say - lora inspired by former Riot worker - Chenbo! Makes amazing characters stylized as splash arts.
v1 - more splash art'y and more creative
v1 high res - less creative, adds more details to characters, higher resolution
Trigger words makes lora much stronger.
Recommended settings:
Sampler:
er_sde
euler
dpmpp_sde_gpu
Scheduler:
beta - for more drawn style
linear_quadratic - for more softened, 3d look
Sampling steps: 25-40
CFG: 4-5
Lora Strength: 0.6-0.8
Description
FAQ
Comments (3)
Does the Krea version work only with the Turbo model, or can it be also used well with the base/raw one as well?
Honestly I only tested the Turbo version, but it technically was trained on raw, hmm..
Great style. Krea version does seem to introduce a bias towards close ups or cowboy shots even when prompting for full body, at least in my testing. Other than that its very nice.