Dolls Kill Collection
Trained with OSTRIS [email protected]
Clubwear: Rude Awakening , Downbeat Spiked, Diatonic, Digital Dreamland, Pitch Strappy,
TimePulse,
Cute: I'm Baby Dress, Shugar Rush ,Polar ,Academic Attitude, Lounge , Boys_Cry
Costume: Nailed it,
Streetwear: Jason, Easy, Stuxnet
Goth: RitualRadiance , WebcamThrills
A collection of outfits from various brands by Dolls Kills.
Description
FAQ
Comments (6)
nice work...
maybe since you are a pro in train, you can save images as PNG and add prompt ;)
My ComfyUI workflow is too complex, so the prompts aren’t read in the metadata by Civit anyway. That’s why I’m posting a prompt template—this way, you can reproduce the clothing.
I add manualy the prompts from first 3 Images ;)
Prompts all custom GPT Prompts ;)
@denrakeiw am I just blind, or where did you post a prompt template for this lora? cant really find it. Btw. why do you need to use BS like "my comfy workflow is too complex for proper metadata" :-D .... cant tell about the meta data reader of civit. - but usually people will download the images and import them into SD or comfyUI to get the prompt... don´t get me wrong. Your loras look dope af, havent tried one so far though. If you´d like to keep your prompts privat, its totally fine. We all know how long it can take to make a prompt finally produce great results! blaming comfy lacking capability to add whatever data of the prompt or workflow you´d like is kinda weird.
Anyway! Dope work, at least based on the images. I ll check the loras out later.
@moelle In images created with ComfyUI, the metadata does not store the prompt but rather the entire workflow.
If you have a relatively complex workflow with multiple samplers, noise blending, and noise injection, the CivitAI metadata reader can no longer process this information properly.
This leaves only two options: either manually entering the prompts, which is too time-consuming for me, or generating much lower-quality sample images using a standard workflow, which doesn’t meet my standards either.
I create my prompts using either GPT-4o or Claude 3.5. It’s not too difficult—I even wrote an article about it and created a custom GPT that is publicly available.
@denrakeiw yeah, I didnt mean to offend you in any way. but so far, I had no issues adding prompt data without the workflow. there are countless custom nodes to handle meta data injection. - but thats probably not news to you. - I´m just not sure how important the civitAI meta data ready is. At least I use to download an image and drop it into comfy or forge to get its prompt. But anyways, its up to you how you handle this stuff. I already had some fun time playing with some of your loras. Its impressive work! Good job!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.



















