
Got tired of not having SDXL equivalents of my favorite LoRAs from SD1.5, so I decided to try and make my own.
Do know that although I incorporated a lot of high-quality images of Tharja/Kawakami-sensei from what I found on the web, I also used some of my own generated images using my favorite LoRAs. (114 images total, repeated 10 times)
Since the two characters were trained using different subfolders, this LoRA effectively has 2 independent trigger words: Tharja (Fire Emblem), and Kawakami Sadayo.
Not recommended to use both triggers at once, unless you plan to make a hybrid or something.
Why two unrelated characters in one LoRA? Because I like both of these characters a lot, and simply haven't seen them on SDXL. Though I'm sure by the time I've realized it, someone who's far better than me would have made a superior LoRA for each of these characters.
Note:
This LoRA doesn't quite function like how my favorite ones do. What I mean by that is, on my favorite LoRAs, I could simply invoke the trigger, and it usually takes care of everything. Here, sure you must use the appropriate trigger to get either character, but it helps to also include more descriptions.
Example prompts:
Tharja:
"Tharja, long black hair, blunt bangs, tiara, body stocking, bridal gauntlets"
Kawakami:
"Kawakami Sadayo, short messy hair, yellow striped shirt, jean skirt"
描述:
You might get away with just using the trigger words on base SDXL, but for checkpoints like SilenteMoney, it's best to be as descriptive as possible for the best outcome.
训练词语: Guinaifen \(Honkai: Star Rail\),Sparkle \(Honkai: Star Rail\)
名称: Double_double_Guinaifen_Sparkle.safetensors
大小 (KB): 223116
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success