Interactive Stable Diffusion
A widgets interface based on ipywidgets library for Stable Diffusion.
Colab
Colab |
Info |
|
Interactive-SD Official |
Note: In this Colab, it is advisable to use the 'Generate in GUI' cell to keep your session active and prevent disconnection issues due to inactivity.
GUI
Features:
Prompt Control:
- Positive Prompt: Guide the model's image generation.
- Negative Prompt: Specify elements to avoid in the generated image.
- Prompt Weights: Emphasize or de-emphasize prompt aspects.
- Long Prompts: Manage lengthy prompts with unlimited characters.
Image Quality Control:
- Steps: Determine denoise steps of image generation.
- CFG: Adjust creativity vs. strict adherence to prompt.
- Sampler: Reduce image noise through denoising.
- Seed: Set seed for reproducibility or use -1 for random.
- Clip Skip: Skip prompt adherence layers.
Style and Adaptation:
- LoRA: Store small modifications to SD model for efficiency.
- Embeddings: Adapt model to a particular style.
- FreeU: Improve diffusion model sample quality at no cost.
- VAE: Choose VAE for different results.
- LCM: Generate images with Latent Consistency Models.
- ControlNet: Enhance context-aware image generation.
- Styles: Add specific styles to generation.
Inpainting and Image Modification:
- Inpaint: Modify images or remove objects using diffusion model.
- Img2img: Modify existing images based on text prompts.
- Adetailer: Automate inpainting with adjustable strength.
Resolution Enhancement:
- High-resolution: Increase image size proportionally.
- Upscaling Tools: Select different models and options for High-resolution.
Additional Points:
- Compatibility: Support for both ckpt and diffusers models, such as
cagliostrolab/animagine-xl-3.0
.
- Model version: SD 1.5 and SDXL diffusers support
- Save images: Automatically save the generated images and specify the saving directory.
- Generation Data: Utilize generation data from previously generated images and configure the parameters in the GUI; currently, this functionality is only available for specific parameters.