| Property | Value |
|---|---|
| License | AGPL-3.0 |
| Base Model | runwayml/stable-diffusion-v1-5 |
| Training Data | 1.2M Instagram images |
| Author | cafeai |
What is cafe-instagram-sd-1-5-v6?
cafe-instagram-sd-1-5-v6 is a specialized Stable Diffusion model fine-tuned for generating Instagram-style Japanese idol and fashion photography. Trained on runwayml/stable-diffusion-v1-5 for approximately 1.6 epochs using 1.2M curated Instagram images, this model leverages BLIP natural language descriptions and booru tags for enhanced image generation capabilities.
Implementation Details
The model employs a sophisticated training approach using various aspect ratios with a base resolution of 768x768 and utilizes the penultimate CLIP layer. For optimal results, it's recommended to use a clip skip of 2 and maintain a resolution of 768x768 or higher.
- Trained on diverse aspect ratios with 768x768 base resolution
- Implements BLIP captioning and booru tag assistance
- Incorporates Instagram hashtags in training data
- Uses penultimate CLIP layer for improved performance
Core Capabilities
- Generation of photorealistic Japanese idol and fashion photography
- Support for various Instagram-style aesthetics
- Enhanced performance with specific prompt structures
- Specialized in generating realistic portraits and fashion shots
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in Instagram-style Japanese idol and fashion photography, utilizing a combination of BLIP descriptions and booru tags for enhanced generation capabilities. Its training on authentic Instagram content makes it particularly effective for creating realistic social media-style images.
Q: What are the recommended use cases?
The model is best suited for generating fashion photography, idol portraits, and Instagram-style content. It's recommended to use the model with a clip skip of 2 and resolution of 768x768 or higher. For optimal results, mixing with other models may enhance performance due to its undertrained nature.