ENGLISH
I created ru551an 1n5t1tut3 as a photorealistic luxurious Russian boarding school background LoRA set in winter, inspired by the refined atmosphere of elegant European adult cinema. It presents an old-world institution shaped by discipline, imperial palace-chic, warm lamplight and cold snowy daylight. The sensuality comes only from the set design, materials and lighting. Every training image is free of people and designed as an empty stage with a wide open center for placing a character.
Inside the set: grand entrance hall, dormitory, classroom, headmistress office, punishment room, dining hall, library, communal showers, gymnasium, teachers lounge and stables.
How to use it: use the trigger word ru551an 1n5t1tut3. Recommended LoRA weight: 0.8-1.0. The concept is intended for SDXL, Illustrious, Pony and natural-language generators such as Krea.
Example prompt: ru551an 1n5t1tut3, photorealistic luxurious Russian boarding school entrance hall in winter, white and gold stucco, dark mahogany paneling, deep green walls, brass, warm amber lamplight, cold snow light through tall arched windows, wide open center floor, cinematic lighting, no humans
Suggested negative prompt: humans, people, body, face, gore, blood, readable text, logo, watermark, low quality, blurry
Tips:
- Start around weight 0.9, then lower it if you want your checkpoint to control more of the architecture.
- Keep the location and camera angle explicit, and mention wide open center when you plan to add a character.
- Combine warm amber interior light with cold snowy daylight to preserve the intended winter identity.
arrssenne
FRENCH
J ai cree ru551an 1n5t1tut3 comme un LoRA photorealiste de backgrounds representant un luxueux pensionnat russe en hiver, inspire par l atmosphere raffinee du cinema adulte europeen elegant. Il presente une institution ancienne marquee par la discipline, le chic des palais imperiaux, la lumiere chaude des lampes et la lumiere froide de la neige. La sensualite vient uniquement du decor, des materiaux et de l eclairage. Toutes les images d entrainement sont sans personne et sont concues comme une scene vide avec un large centre degage pour placer un personnage.
Le set contient : grand hall d entree, dortoir, salle de classe, bureau de la directrice, salle de discipline, salle a manger, bibliotheque, douches communes, gymnase, salon des professeurs et ecuries.
Utilisation : employez le trigger word ru551an 1n5t1tut3. Poids LoRA recommande : 0.8-1.0. Le concept est destine a SDXL, Illustrious, Pony et aux generateurs en langage naturel comme Krea.
Exemple de prompt : ru551an 1n5t1tut3, photorealistic luxurious Russian boarding school entrance hall in winter, white and gold stucco, dark mahogany paneling, deep green walls, brass, warm amber lamplight, cold snow light through tall arched windows, wide open center floor, cinematic lighting, no humans
Negative prompt suggere : humans, people, body, face, gore, blood, readable text, logo, watermark, low quality, blurry
Conseils :
- Commencez avec un poids proche de 0.9, puis reduisez-le si vous voulez laisser davantage de controle a votre checkpoint.
- Precisez le lieu et l angle de camera, et ajoutez wide open center si vous voulez placer un personnage.
- Combinez une lumiere interieure ambree avec une lumiere froide de neige pour conserver l identite hivernale.
arrssenne
Description
FAQ
Comments (6)
What is the reason you aren't training a FLUX LoRA?
As for Flux, the first issue is that I honestly have no idea which model I should use to generate the preview images afterward. So if you have a particular model you’d recommend, I’d definitely be interested in trying it.
Now that I’ve switched to ComfyUI for my Krea workflow, I could also experiment with setting up a Flux workflow and see how well it works for me.
The main reason I haven’t really considered Flux so far is that there hasn’t been much demand for it from the people who follow my work. But if people start asking for Flux versions, and if the conversion cost on Civitai is reasonable, I’d definitely consider adding Flux to the models I support.
Thanks for this. Release datasets for blue buzz is a good idea!
Thank you! I'm really happy to see that my work is appreciated!
My LoRAs will always be free. Personally, I don't like the idea of charging people for LoRAs.
However, datasets are a different story, since I've invested quite a bit of money into creating them, and they also serve as the foundation for my own training.
Once you've purchased a dataset, you're completely free to use it however you want. You can train your own models for Flux, Anima, or anything else. I honestly don't mind at all! Once you've paid for the dataset, what you do with it is entirely up to you.
At least this way, you'll have access to a beautiful, high-quality collection of images ready for training!
@arrssenne By "creating", do you mean to say all the images therein were generated?
@firemanbrakeneck Yes, that’s exactly what I mean! All the images in my datasets are AI-generated.
I use Kimi to help develop detailed prompts that guide ChatGPT in generating the images. Both services cost money, so creating these datasets involves expenses as well as the time spent designing the scenes, refining the prompts, and preparing the images for training.
That’s why I charge for the datasets, while my LoRAs will always remain free. 🙂