If you’d like to show some love, feel free to react w/ emoji, post, rate, follow, or simply drop some feedback, I'd appreciate it. 🫡🫰 Feel free to also report any issues.
Anima style is softer, seems to properly balance colors and shading away from some scenes. Freckles should also be reduced, did a better job at handling them. Took quite a bit to get right or at least to how I like it. The dataset was updated, though I removed more than I added.
Anima Base v1.0
Trained on Anima - base-v1.0, images also generated on base. Generation details, prompts, and model info are listed with most images: CFG: 5, Steps: 20-30, Sampler: DPMPP 2M Simple, and SeedVR2. Feel free to experiment!
Trigger Word: @mxpln
Illustrious trigger words are on their detail card: mxpln. Eye problems/smudge happen way less often in Anima. It handles off styles better. If you can't use ADetailer, try Anima.
Description
Adjusted prompts.
Less image blur/smudge
Removed older images and added newer ones.
Now closer to the latest artist’s style.
Should reduce some issues present with v2, at least to a certain degree
Mixplin Base/Hires/Hires+ADetailer
Simple Pose
https://imgsli.com/NDI1Mzg5/0/2
Camera distance/Complex Poses
FAQ
Comments (12)
Like this lora. But the belly protrusion looks a bit odd. Is there anynegative words I could add?
Since I don't think there really is a specific Danbooru tag tied to the concept, you can try countering it with words the AI might confuse it for; "belly, plump, big belly, linea alba, abdominal muscles". These can sometimes help tone down. Here's an example.
@Ty_Lee Thank you so much! I'll give it a try right away.
nice
Insane how well this LoRA conforms to complex prompts (~625 tokens). Even the poorest of generations don't look too bad and only require a bit of touch up in in-painting or Photoshop. Quite possibly the most plug-and-play LoRA on here and I've used HUNDREDS. Incredible work!
I appreciate it 😊 I’m really glad the LoRA is working well for you. I did try to minimize issues with the eyes and face, training the base model and certain styles can sometimes clash and have these issues. Thanks!
@Ty_Lee Interestingly enough, I've never had any issues mixing your LoRAs together (Cherry-gig and this one) regarding the eyes and face. They're very consistent.
greate
what is clip skip? is there a node for it? and do i need to use Load VAE node?
Sorry for the late reply, you probably figured it out already you. Don't actually need clip skip if you are using ComfyUI. Also, don't worry about vae, it's built in. Only have to connect them from checkpoint.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.