DiffSynth is a new Diffusion engine. We have restructured architectures including Text Encoder, UNet, VAE, among others, maintaining compatibility with models from the open-source community while enhancing computational performance. This version is currently in its initial stage, supporting SD and SDXL architectures. In the future, we plan to develop more interesting features based on this new codebase.
Create Python environment:
conda env create -f environment.yml
We find that sometimes conda cannot install cupy correctly, please install it manually. See this document for more details.
Enter the Python environment:
conda activate DiffSynthStudio
python -m streamlit run DiffSynth_Studio.py
sdxl_turbo_ui.mp4
We can generate images with very high resolution. Please see examples/sd_text_to_image.py for more details.
| 512*512 | 1024*1024 | 2048*2048 | 4096*4096 |
|---|---|---|---|
 |
 |
 |
 |
Generate images with Stable Diffusion XL. Please see examples/sdxl_text_to_image.py for more details.
| 1024*1024 | 2048*2048 |
|---|---|
 |
 |
Generate images with Stable Diffusion XL Turbo. You can see examples/sdxl_turbo.py for more details, but we highly recommend you to use it in the WebUI.
| "black car" | "red car" |
|---|---|
 |
 |
This example is implemented based on Diffutoon. This approach is adept for rendering high-resoluton videos with rapid motion. You can easily modify the parameters in the config dict. See examples/diffutoon_toon_shading.py. We also provide an example on Colab.
Diffutoon.mp4
This example is implemented based on Diffutoon, supporting video editing signals. See examples\diffutoon_toon_shading_with_editing_signals.py. The editing feature is also supported in the Colab example.
Diffutoon_edit.mp4
This example is provided for developers. If you don't want to use the config to manage parameters, you can see examples/sd_toon_shading.py to learn how to use it in native Python code.
toon_dancing_girl.mp4
Given a prompt, DiffSynth Studio can generate a video using a Stable Diffusion model and an AnimateDiff model. We can break the limitation of number of frames! See examples/sd_text_to_video.py.
lightning.mp4
We provide an example for video stylization. In this pipeline, the rendered video is completely different from the original video, thus we need a powerful deflickering algorithm. We use FastBlend to implement the deflickering module. Please see examples/sd_video_rerender.py for more details.
winter_stone.mp4
If you are not native English user, we provide translation service for you. Our prompter can translate other language to English and refine it using "BeautifulPrompt" models. Please see examples/sd_prompt_refining.py for more details.
Prompt: "一个漂亮的女孩". The translation model will translate it to English.
| seed=0 | seed=1 | seed=2 | seed=3 |
|---|---|---|---|
 |
 |
 |
 |
Prompt: "一个漂亮的女孩". The translation model will translate it to English. Then the refining model will refine the translated prompt for better visual quality.
| seed=0 | seed=1 | seed=2 | seed=3 |
|---|---|---|---|
 |
 |
 |
 |