Using Image Models
Stable Diffusion is currently the most popular model in the field of image content generation, and almost all popular models in the community are derived from it.
stable-diffusion-webui is currently the most popular image generation workstation software, and we recommend using it to use image models.
Currently, the common usage process is:
- Users download the model to their local computer.
- Move the model file to the directory specified by stable-diffusion-webui.
- Start stable-diffusion-webui to use the model.
For usage examples of other software, please refer to the corresponding official documentation.
| Name | Advantages | Disadvantages |
|---|---|---|
| InvokeAI | Easier installation, better usability, better UI integration | Fewer advanced features compared to webui |
| ComfyUI | Supports node orchestration, more convenient customization | Fewer advanced features compared to webui |
| StableStudio | Official open source software of Stability.AI, supports Stable Diffusion XL model, front-end and back-end separation architecture | Currently newly open source, weak ecosystem |
For developers, it is recommended to use the huggingface diffusers package, which allows for better control through Python code. However, currently (as of April 27, 2023), diffusers does not support dynamically loading fine-tuned models or using multiple fine-tuned models simultaneously. Additionally, there are differences in the model format supported by diffusers and stable-diffusion-webui, and scripts are needed for conversion.