Changelog:
V2.2_pynoise: Trained using DAdaptLion optimizer, with minimal SNR gamma enabled. Smaller network_dim/network_alpha (32/16), optimized size of model (36.12M), and reduced overfitting.
V1.0: Initial release. Model with multi-resolution noise.
变更日志：
V2.2_pynoise：使用DAdaptLion优化器训练，启用最小信噪比伽马值。减小了dim/alpha值（32/16）和文件体积（36.12M），减轻了过拟合现象
V1.0_pynoise：首发版本，添加添加多分辨率噪声
————————————————————————————————————————
LoRA model of Mirai Kasuga (春日未来) from media-mix project THE IDOLM@STER MILLION LIVE! (アイドルマスターミリオンライブ！).
The trigger word is mirai kasuga (million live). It's well known that hairclip is Mirai-chan's most notable feature, but I notice that many official illustrations don't contain it, perhaps meaning that the hairclip may be optional. When using this model, you may need add hairclip (or a specific type of hairclip e.g. pink hairclip/x hairclip/star hairclip) tag into your prompt if you want its appearance in your illustration. Besides, you can use tags such as hair scrunchie/hair bow/hair ornament/hair flower to adjust the item for holding Mirai-chan's side ponytail in place.
The LoRA weight of samples is 0.85; you can also try other weights. The greater the weight, the closer the character style is to the game card of THE IDOLM@STER MILLION LIVE! THEATER DAYS (ミリシタ) and the new tachie drawn after the release of the game.
Note: I load my LoRA model through the additional network WebUI plugin, so the prompt of sample illustrations don't contain <lora:XXX:X> tag.
I'll appreciate it if you're willing to use this LoRA model and share AI arts to other people. My models and illustrations are free for everyone, and if you like them, please support me at Patreon or 爱发电.
————————————
跨媒体企划《偶像大师 百万现场》的角色春日未来的LoRA模型
模型的触发词为 mirai kasuga (million live)。众所周知，发卡是未来这孩子最显著的特征，称之为本体也不为过。但是我发现有相当数量的官方图并没有包含发卡，或许意味着在官方看来发卡是可选项。在使用本模型时，如果您需要发卡稳定出现在画面中，就需要加入hairclip（或者pink hairclip/x hairclip/star hairclip等特定样式的发卡）tag。另外，您也可以使用hair scrunchie/hair bow/hair ornament/hair flower这些tag调整未来用来固定侧单马尾的物件
示例图的LoRA权重为0.85，您也可以尝试其他权重值。权重越大，角色风格越接近游戏《偶像大师 百万现场 剧场时光》（MLTD）游戏卡面以及该游戏发行后新绘制的立绘
我使用的是WebUI的可选附加网络（additional network）插件，所以示例图的prompt里并没有<lora:XXX:X>的tag，还请注意
如果您愿意使用本LoRA模型并将作品分享给其他人，本人将不胜感激。本人制作的所有模型和绘画皆免费提供，如果觉得还不错的话，欢迎在Patreon或爱发电上支持我，非常感谢。
Description
V1.0: Initial release. Model with multi-resolution noise.
FAQ
Comments (9)
This is an amazing job, perfection all over the place, excellent, no way to top this one!
next, Momoko or Iku, please!!
If time permits, I will finish all character models of MILLION LIVE! step by step, but it's more according to my interests, so I am not sure what the next character will be. You may need to wait a little longer.
An excellent creator has made a 50+-in-one multi-subject character model of MILLION LIVE!, and maybe it can meet your requirement. Try it out.
@KaorinP_plkm160 I've tried that 50 character LORA, not to disrespect the work that creator did and sound ungrateful but it was a tall task and I think a bit too much to pack into one LORA. You figure it'd be convenient but you're basically packing so much data into one thing it feels like tags bleed over each other due to how many characters there are with overlapping attributes like hair length, hair color, eye color, etc.
It's why I prefer singular character LORAs even if it means it'll take much longer for creators like yourself more time to create them.
I'd kill for a Momoko LORA at this point too. I've seen more than one pixiv creator/artist post up their own custom LORA of Momoko and I've tried asking nicely if they'd ever release it to the public but usually get no reply. One person was kind enough to at least reply to my request and say they're keeping theirs private at least.
@a22681 THIS, yes, so much this, also: the amount of character generated with suddenly incorrect eye color is almost comical
@DERESUTE Why do you have the same name as an abandoned Idolmaster LORA on huggingface btw? Are you the creator of that LORA even if your civitai account here has no uploads? Or just a fan?
@DERESUTE @DERESUTE Sorry for the late reply. I'm trying making the Momoko's character model (some materials have been prepared and processed before so it's not a tough job). According to my experience, AI may have difficulty in learning Momoko's little ahoge, so it may take time for me to adjust the model.
@KaorinP_plkm160 YESS! thank you ,thank you, thank you!! You did an amazing job with Anna's ahoge so coming from you i am sure it will end being really accurate, please do it at your own peace!! !! also forgive me i if was an annoyance for just asking for stuff without being of help for others m( . _ . )m...
@DERESUTE Apologies if that came off rude, just wondering. I find it too coincidental that your username and your enthusiasm for the same subject matter (idolmaster girls) is exactly the same to that of a LORA of literally - idolmaster girls.
@a22681 i replied to you 2 times and ended deleting both because my english is bad. i am the one at fault here, sorry. Yeah, it's just a coincidence.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.