@VXP
Hálainnithomiinae
VXP Anima Test
As stated in my comment on this model's page: I'm loving the model's overall style so far, including across differing styles and levels of realism, but it seems to consistently add unprompted details (excessive jewelry, clothing accents, etc.) that require actively combating with prompting and negatives to keep them in check. The default negatives and overall parameters could, likewise, use some tuning. <- As you can see, I used less steps and a lower CFG and achieved arguably better image quality than the defaults, and these results also apply without the LoRAs used in this image (which I pretty much only used to give the hands a boost). I'll be ecstatic to see a turbo/hyper version of this model, and it obviously works well with the Anima Turbo LoRA and the Lenovo UltraReal LoRA. Unrelated side-note: Yay, they finally added "golden eyes" to the autotags! Don't know how long that's been a thing, but I'm happy to see it.
Show Parameters
- Size
- 1024 x 573
- Sampling Steps
- 13
- Sampling Method
- Euler
- CFG Scale
- 3.3
- Seed
- 5007123665743727
- Negative
- (child, young:2), unclear eyes, (low quality, worst quality:3.14), (censored, censoring, pixelated:2), (text, logo, watermark, patreon:1.2), CG, blendr, render, ai-generated, CGI, fake, (bad proportions, long neck, long torso, long fingers:1.4), large breasts, (underexposed, unnatural lighting:1.6), (bad hands, bad fingers, extra digit, excess digits, extra fingers:1.5), (see-through ears, translucent ears:1.3), (big eyes, small face:1.6),
