You can deploy this model and start earning money today!

mushroomfleet/
WAN14b-CyberS0C13TYv77

Name: WAN14b-CyberS0C13TYv77
Availability: InStock

mit

Views

No views yet

Cybersociety V77

A collection of LoRA models for generating Cybersociety V77 images, trained on WAN 14B 2.1 and optimized for use with WAN 14B 2.2.

📁 Repository Structure

├── cyber-s0c13ty-v0b-adamw8bit.safetensors
├── wan-cybersoc-v77-945.safetensors
├── wan-cybersoc-v77-1428.safetensors
├── captions/
│   ├── [matching filename structure text files]
└── ENH/
    ├── [system prompt .md files]

🎯 LoRA Models

🥇 wan-cybersoc-v77-945.safetensors (Recommended)

Primary style model for Cybersociety V77
Training dimensions: 945 (recommended for 1024x image datasets)
Quality: Achieved superior results, surpassing the 632 sized version
Best for: General cyberpunk style generation

🔬 wan-cybersoc-v77-1428.safetensors (Experimental)

High-resolution training experiment
Note: Dataset was not 2K dimensions despite hires training
Best for: Experimental high-resolution outputs

👤 cyber-s0c13ty-v0b-adamw8bit.safetensors (Character Focus)

Character LoRA trained specifically on "S0C13TY cyber-samurai"
Dataset: Smaller, focused character dataset
Best for: Character-specific cyber-samurai generation

📝 Additional Resources

Captions Folder

Contains text files with matching filename structure for each LoRA, providing detailed caption data used during training.

ENH Folder (Enhancement Prompts)

Contains system prompt .md files that transform simple or complex prompts into well-aligned prompts optimized for the Cybersociety style.

Usage:

Recommended: Create your own prompts using the base template provided
Originally designed for: Qwen25VL workflow integration
Alternative use: Can be used in chat interfaces if you cannot run a local LLM
Implementation: Use each .md file as a complete system prompt

🔧 Technical Notes

Base Model: WAN 14B 2.1 (training)
Target Model: WAN 14B 2.2 (inference)
Dimension Numbers: Relate to codec safe dimensions in video processing
Recommended Training Dimension: 960 for optimal results with 1024x datasets

🚀 Getting Started

Download the recommended wan-cybersoc-v77-945.safetensors model
Load with WAN 14B 2.2 for image generation
Use system prompts from the ENH folder to optimize your prompts
Experiment with the character-focused v0b model for cyber-samurai themes

📋 Recommendations

Start with the 945 model for best overall quality
Leverage the ENH system prompts for optimal style alignment
Refer to captions folder for training data insights

Cybersociety V77 - Bringing the future of cyberpunk aesthetics to your image generation workflow.

mushroomfleet/WAN14b-CyberS0C13TYv77