Weight: 1.0
Garreg Mach Monastery Uniform: long hair, pink hair, blunt bangs, twintails, pink eyes, garreg mach monastery uniform, short dress, black dress, buttons, brown belt, white shirt, white ascot, sleeves rolled up, blue thighhighs
Timeskip: long hair, pink hair, ponytail, pink eyes, short dress, black dress, sleeveless dress, cleavage cutout, pink sleeves, puffy detached sleeves, vertical-striped sleeves, red gloves, underbust, red corset, red belt, pink skirt, showgirl skirt, red thighhighs, thighhighs under boots, black footwear, thigh boots, hoop earrings, bare shoulders
Three Hopes: long hair, pink hair, blunt bangs, ponytail, single hair bun, pink eyes, red capelet, red jacket, open jacket, puffy short sleeves, short dress, pink dress, red bow, red gloves, underbust, black corset, cross-laced legwear, red thighhighs, hoop earrings, cleavage
Holiday Layabout: long hair, pink hair, blunt bangs, twintails, hair flower, white flower, red headwear, santa hat, pink eyes, santa costume, fur trim, short dress, red dress, shoulder cutout, fur collar, sleeves past elbows, white gloves, pink thighhighs, vertical-striped thighhighs, bare shoulders
Deer's Two-Piece: long hair, pink hair, blunt bangs, twintails, sunglasses, eyewear on head, pink eyes, pink bikini, off-shoulder bikini, vertical-striped bikini, bare shoulders, collarbone, cleavage, large breasts
The results of your generations are dependent on the additional LoRAs, weights, and models you use, so it may not work or come out as consistent as my sample images.
Description
Initial release.
FAQ
Comments (5)
Sir, I love your LoRAs, but there’s a little problem with the way the prompts are constructed.
Let me explain: when you try to add two or more characters to the scene, they tend to merge.
It’s not an AI or Anima problem. It’s because the LoRA triggers don’t associate the character’s name with the character they were trained on. So, if you add “1girl” and “1girl,” both of them are interpreted as the same character.
The correct way to train the LoRA should be:
Character name, world or franchise, clothing or theme, character description, clothing description.
That way, you could use a bunch of different LoRAs in the same scene, and Anima could resolve the final image much more precisely.
I hope this can be helpful. I know you didn’t ask for advice, but I just wanted to mention it because I think there’s some room for improvement.
I've gotten this suggestion before Anima came out, but having no trigger is a deliberate choice. Now I would like to throw in images with multiple characters in it from time to time, but SwarmUI is so frustratingly slow when it comes to regional prompting. I can't be bothered to open up Forge or get a ComfyUI workflow to make those kinds of images. Instead, I asked someone to make them for me as a test, which should be visible in the gallery below.
Appreciate the suggestions from both of you but I'm not convinced enough to go back to triggers again.
@justTNP I get the approach with regional prompting, especially with the Anima Cosmos regional prompting node. But in reality, most of the time, if you add a natural-language description to your prompt, Anima’s Qwen 0.6B CLIP is pretty capable of rendering exactly what you ask for.
So yeah, regional prompting definitely works, no doubt about that, since you can assign a different LoRA to each colored section. But it requires extra nodes and a more complicated workflow to make it work.
I understand that your LoRAs are more focused on that regional-prompting approach and on rendering a single character accurately, though.
You have such an excellent set of loras, but they really only work best if you are only going to do single characters or pairing them with generic characters. Character base trigger, and even outfit triggers would go a long way.
First unit that died during my playthrough o7.