Ai2Ai_v1
🌟 Model Details
- Model Name: Ai2Ai_v1
- Description: A LoRA trained on AI-generated images sourced from Pinterest.
- Version: v1
- Release Date: I don't remember.
- Framework: Kohya Trainer Colab
🌍 Intended Use
This LoRA is designed for generating Pinterest girls(?). Good-ish at facial features and hats(?). it's not optimized for rendering full body and clothing designs.
📊 Performance
- Accuracy: High on face and hat generation, lower on body details
- Precision: High on face and hat generation, lower on body details
⚡ Limitations and Bias
- Limitations: This LoRA model is relatively small, resulting in the generation of somewhat identical faces.
- Bias Considerations: Beware of potential artifacts such as "jpeg artifacts," "cheek bulge," and the model's potential to be overly powerful in certain aspects. Put these in negative prompts.
🎯 Training Data
- Data Sources: Images from Pinterest
- Data Size: 82 images, approximately 12 MB
- Data Preprocessing: AI-generated images, tagged with wdtagger