Sign In

Russian Ministry of Justice/MVD Parade student uniform

3

Updated: Sep 30, 2026

clothing

Not loaded

Download

1 variant available

bf16 SafeTensor

Russian_Ministry_of_Justice_and_MVD_parade_student_uniform_krea2_3371803_epoch_5.safetensors

BF16, good balance • 218 MB

Verified:

Type
LoRA
Stats

149

3

6

Generation

Not loaded

Reviews
Published

Sep 30, 2026

Base Model

Krea 2

Training
Steps: 4,000
Epochs: 10
Usage Tips
Strength: 0.8
Hash
AutoV2
1AFD96D760
default creator card background decoration
Reactions - 211

211

Followers - 6

6

Generations - 11

11

Black diamond
ComfyUI_00052_.png

Justice Dress v2 — Ministry of Internal Affairs Uniform (Mix & Customization)

A Krea 2 LoRA trained to reproduce the formal uniform dress worn by personnel and female students of the Russian Ministry of Internal Affairs (МВД). Unlike my first model, this LoRA is designed as a modular clothing constructor. It allows you to flexibly mix, match, or completely exclude specific elements of the uniform in your prompts, generating various combinations such as: skirt + shirt, classic trousers + shirt + tie, or full formal dress tunic dress code.

The model successfully separates and reproduces:

  • The upper layer: A single-breasted formal dress tunic with polished gold buttons and distinct blue piping on the collar and cuffs.

  • The shirts: Long-sleeved and short-sleeved formal shirts in both crisp white and classic light blue colors.

  • The lower layer: A tight-fitting corporate pencil skirt or classic straight-leg trousers with thin blue stripes/piping down the side seams.

  • Accessories: A personnel-issue wide tie featuring the Russian coat of arms and gold tie clips.

The LoRA was trained on 81 high-quality reference images (including various combinations of clothing, full-body shots, close-ups, and flat-lay product photos) using Krea 2 Base and advanced training workflows. It is designed to strictly separate the concepts so that the AI doesn't "glue" the clothes together, allowing total freedom in characters, poses, environments, and lighting conditions.

Sometimes, slight anomalies on small details like the gold eagle emblem on the tie may appear due to pixel limitations on wide shots; I recommend touching those up using FaceDetailer/Inpaint or doing a high-res upscale.

Trigger words & Clothing Constructor Tokens:

To achieve the best results with Krea 2's natural language processing, use the main style anchor followed by the exact descriptive tokens for the clothes you want to see:

  • Main style anchor: ruUniformStyle

  • For shirts: light blue long-sleeved formal shirt for Ministry of Internal Affairs personnel, white short-sleeved formal shirt for Ministry of Internal Affairs personnel, light blue short-sleeved formal shirt for Ministry of Internal Affairs personnel, white long-sleeved formal shirt for Ministry of Internal Affairs personnel

  • For the jacket: formal dress tunic of the Ministry of Internal Affairs (you can specify open jacket or fully closed jacket)

  • For the bottom: women's skirt for Ministry of Internal Affairs personnel or classic trousers with blue stripes for Ministry of Internal Affairs personnel

  • For the tie: Ministry of Internal Affairs personnel-issue tie featuring the Russian coat of arms

Recommended usage:

Describe your character, pose, and environment normally using natural language, activate the main style, and list the exact clothing pieces for the mix.

Example (Skirt + Shirt combination):

ruUniformStyle, a young woman standing in a classroom near a window, wearing a white short-sleeved formal shirt for Ministry of Internal Affairs personnel, a Ministry of Internal Affairs personnel-issue tie featuring the Russian coat of arms, and a dark women's skirt for Ministry of Internal Affairs personnel, realistic photography, natural lighting

Training details

  • Base model: Krea 2 Base

  • Training images: 81

  • Training steps: 4000 (with 10 intermediate checkpoint saves)

  • Resolution: 1024

  • Batch size: 2

  • LoRA rank (Dim): 32

  • LoRA alpha: 16

  • Optimizer: Prodigy (Adaptive Learning Rate)

  • Shuffle Captions: Enabled (Keep Tokens: 1)

Intended use

This LoRA was created as an experimental educational project for studying generative AI, advanced multi-concept dataset structure, and text encoder token separation. The model is intended to reproduce the uniform structure and clothing variety itself, not any particular real person.