| Property | Value |
|---|---|
| License | CC0 1.0 |
| Language | English |
| Token Count | 2 tokens |
| Training Steps | 15000 steps |
What is bad-artist?
bad-artist is a specialized textual-inversion embedding model designed for use in negative prompts with Stable Diffusion. It comes in two variants: the standard 'bad-artist' version which produces unique images, and 'bad-artist-anime' which is optimized for anime-style generations. The model was trained with 2 vectors per token for 15,000 steps at 500x500 resolution on an Anything-v3-based model.
Implementation Details
The embedding is specifically designed to work with the 'by' prefix in negative prompts, such as "sketch by bad-artist" or "painting by bad-artist". This implementation helps define what the model should avoid generating, resulting in higher quality outputs.
- Uses only 2 tokens for efficient processing
- Trained on Anything-v3 base model
- Optimized for 500x500 resolution
- Compatible with various Stable Diffusion workflows
Core Capabilities
- Improves image quality through negative prompting
- Offers two specialized versions for different art styles
- Works effectively with minimal prompt engineering
- Maintains high quality at various image resolutions
Frequently Asked Questions
Q: What makes this model unique?
This model's strength lies in its ability to improve image generation through negative prompting, using just two tokens while offering both standard and anime-optimized versions.
Q: What are the recommended use cases?
The model is best used in negative prompts with the 'by' prefix, such as "sketch by bad-artist" or "photograph by bad-artist", to help guide the generation process away from undesired artistic styles.