Anyone tried the Nya-Iris Anima checkpoint? Drop your recommended settings.

The Forge / 工坊 Models / LoRA / 模型与 LoRA
1

Just found Nya-Iris’s Anima on Civitai. The author says 1.0 is out but this is still Preview3, kinda confusing. I’m an illustrator looking for a solid anime base model, so I’m gonna jot down the settings and give it a shot.

Here’s what the author recommends:
Steps 12 to 40, CFG 5 to 6, sampler EulerA or dpmpp_2m_sde_gpu, scheduler SGM_Uniform / Simple / Beta.

Also, the author specifically says no need for positive quality prompts anymore, but negative ones are still a must. That’s a nice time-saver—used to be a pain stacking up those long positive prompt strings.

Anyone who’s actually run it, how’s the image style? I’m mainly curious about its thick paint and line art performance.

2

Yeah, I’ve noticed more and more new models are moving away from stacking a bunch of positive quality keywords.

3

CFG 5-6 is kinda low, won’t the output look too plain?

5

Preview3 is still using the 1.0 label, the versioning is a bit of a mess lol.

6

Only 12 steps and it works? That’s pretty fast for generating images.

7

The anime base models have been getting crazy competitive lately.

8

Low CFG doesn’t equal flat/plain results, depends on the base model itself, with newer models like this low CFG actually makes it less prone to overexposure