TrinArt Stable Diffusion v2
| Property | Value |
|---|---|
| Author | naclbit |
| License | CreativeML OpenRAIL-M |
| Training Hardware | 8x NVIDIA A100 40GB |
| Framework | Diffusers |
What is trinart_stable_diffusion_v2?
TrinArt Stable Diffusion v2 is an advanced fine-tuned version of Stable Diffusion specifically optimized for anime and manga-style image generation. The model has been trained on approximately 40,000 carefully curated high-resolution anime/manga-style images over 8 epochs, incorporating advanced techniques like dropouts and improved tagging strategies.
Implementation Details
The model features three distinct checkpoint versions (60k, 95k, and 115k steps) and is implemented using the Diffusers library. It was trained using a custom dataset loader with specific augmentations including XFlip, center crop, and aspect-ratio locked scaling, utilizing a learning rate of 1.0e-5 and 10% dropouts.
- Multiple checkpoint options for different style intensities
- Supports both Text-to-Image and Image-to-Image generation
- Compatible with Diffusers pipeline integration
- Optimized for both performance and quality
Core Capabilities
- High-quality anime and manga style image generation
- Maintains original SD aesthetics while adding anime influence
- Supports various inference methods including Gradio web UI
- Flexible deployment options with different checkpoint versions
Frequently Asked Questions
Q: What makes this model unique?
This model stands out by carefully balancing the original Stable Diffusion aesthetics with anime/manga styling, offering three different checkpoint versions to allow users to choose their preferred style intensity. It's not the 19.2M images Character Model but an improved version of the original Trin-sama Twitter bot model.
Q: What are the recommended use cases?
The model is ideal for generating anime/manga-style artwork, character designs, and artistic compositions. It's particularly effective for both text-to-image and image-to-image transformations, with special consideration for maintaining artistic consistency.
