DiffSynth-Engine

DiffSynth-Engine is a high-performance engine geared towards buidling efficient inference pipelines for diffusion models.

Key Features:

• Thoughtfully-Designed Implementation: We carefully re-implemented key components in Diffusion pipelines, such as sampler and scheduler, without introducing external dependencies on libraries like k-diffusion, ldm, or sgm.

• Extensive Model Support: Compatible with popular formats (e.g., CivitAI) of base models and LoRA models , catering to diverse use cases.

• Versatile Resource Management: Comprehensive support for varous model quantization (e.g., FP8, INT8) and offloading strategies, enabling loading of larger diffusion models (e.g., Flux.1 Dev) on limited hardware budget of GPU memory.

• Optimized Performance: Carefully-crafted inference pipeline to achieve fast generation across various hardware environments.

• Cross-Platform Support: Runnable on Windows, macOS (Apple Silicon), and Linux, ensuring a smooth experience across different operating systems.

Quick Start

Requirements

• Python 3.10+
• NVIDIA GPU with compute capability 8.6+ (e.g., RTX 50 Series, RTX 40 Series, RTX 30 Series. Please see here for more details about your GPUs.) or Apple Silicon M-series.

Installation

Install released version (from PyPI):

pip3 install diffsynth-engine

Install from source:

git clone https://github.com/modelscope/diffsynth-engine.git && cd diffsynth-engine
pip3 install -e .

Usage

Text to image

from diffsynth_engine import fetch_model, FluxImagePipeline

model_path = fetch_model("muse/flux-with-vae", path="flux_with_vae.safetensors")
pipe = FluxImagePipeline.from_pretrained(model_path, device='cuda:0')
image = pipe(prompt="a cat")
image.save("image.png")

Text to image with LoRA

from diffsynth_engine import fetch_model, FluxImagePipeline

model_path = fetch_model("muse/flux-with-vae", path="flux_with_vae.safetensors")
lora_path = fetch_model("DonRat/MAJICFLUS_SuperChinesestyleheongsam", path="麦橘超国风旗袍.safetensors")

pipe = FluxImagePipeline.from_pretrained(model_path, device='cuda:0')
pipe.load_lora(path=lora_path, scale=1.0)
image = pipe(prompt="a girl, qipao")
image.save("image.png")

For more details, please refer to our tutorials (English, 中文).

Showcase

Contact

If you have any questions or feedback, please scan the QR code below, or send email to muse@alibaba-inc.com.

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

Citation

If you use this codebase, or otherwise found our work helpful, please cite:

@misc{diffsynth-engine2025,
      title={DiffSynth-Engine: a high-performance diffusion inference engine},
      author={Zhipeng Di, Guoxuan Zhu, Zhongjie Duan, Zihao Chu, Yingda Chen, Weiyi Lu},
      year={2025},
      publisher = {GitHub},
      howpublished = {\url{https://github.com/modelscope/diffsynth-engine}},
}