Tianrui Feng^*,1, Zhi Li¹, Shuo Yang¹, Haocheng Xi¹, Muyang Li², Xiuyu Li¹, Lvmin Zhang³, Keting Yang⁴, Kelly Peng⁵, Song Han², Maneesh Agrawala³, Kurt Keutzer¹, Akio Kodaira¹, Chenfeng Xu^†,1,6

¹UC Berkeley   ²MIT   ³Stanford University  ⁴Independent Researcher ⁵First Intelligence   ⁶UT Austin

^† Project lead, corresponding to xuchenfeng@berkeley.edu

^* Work done when Tianrui Feng was a visiting student at UC Berkeley advised by Chenfeng Xu.

StreamDiffusionV2 is an open-source interactive diffusion pipeline for real-time streaming applications. It scales across diverse GPU setups, supports flexible denoising steps, and delivers high FPS for creators and platforms. Further details are available on our project homepage.

• [2025-11-10] 🚀 We have released our paper at arXiv. Check it for more details!
• [2025-10-18] Release our model checkpoint on huggingface.
• [2025-10-06] 🔥 Our StreamDiffusionV2 is publicly released! Check our project homepage for more details.

• OS: Linux with NVIDIA GPU
• CUDA-compatible GPU and drivers

conda create -n stream python=3.10.0
conda activate stream
# Require CUDA 12.4 or above, please check via `nvcc -V`
pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu124
pip install -r requirements.txt 
python setup.py develop

# 1.3B Model
huggingface-cli download --resume-download Wan-AI/Wan2.1-T2V-1.3B --local-dir wan_models/Wan2.1-T2V-1.3B
huggingface-cli download --resume-download jerryfeng/StreamDiffusionV2 --local-dir ./ckpts --include "wan_causal_dmd_v2v/*"

# 14B Model
huggingface-cli download --resume-download Wan-AI/Wan2.1-T2V-14B --local-dir wan_models/Wan2.1-T2V-14B
huggingface-cli download --resume-download jerryfeng/StreamDiffusionV2 --local-dir ./ckpts --include "wan_causal_dmd_v2v_14b/*"

We use the 14B model from CausVid-Plus for offline inference demo.

python streamv2v/inference.py \
--config_path configs/wan_causal_dmd_v2v.yaml \
--checkpoint_folder ckpts/wan_causal_dmd_v2v \
--output_folder outputs/ \
--prompt_file_path prompt.txt \
--video_path original.mp4 \
--height 480 \
--width 832 \
--fps 16 \
--step 2

Note: --step sets how many denoising steps are used during inference.

torchrun --nproc_per_node=2 --master_port=29501 streamv2v/inference_pipe.py \
--config_path configs/wan_causal_dmd_v2v.yaml \
--checkpoint_folder ckpts/wan_causal_dmd_v2v \
--output_folder outputs/ \
--prompt_file_path prompt.txt \
--video_path original.mp4 \
--height 480 \
--width 832 \
--fps 16 \
--step 2
# --schedule_block  # optional: enable block scheduling

Note: --step sets how many denoising steps are used during inference. Enabling --schedule_block can provide optimal throughput.

Adjust --nproc_per_node to your GPU count. For different resolutions or FPS, change --height, --width, and --fps accordingly.

A minimal web demo is available under demo/. For setup and startup, please refer to demo.

• Access in a browser after startup: http://0.0.0.0:7860 or http://localhost:7860

• Demo and inference pipeline.
• Dynamic scheduler for various workload.
• Training code.
• FP8 support.
• TensorRT support.

StreamDiffusionV2 is inspired by the prior works StreamDiffusion and StreamV2V. Our Causal DiT builds upon CausVid, and the rolling KV cache design is inspired by Self-Forcing.

We are grateful to the team members of StreamDiffusion for their support. We also thank First Intelligence and Daydream team for their great feedback.

We also especially thank DayDream team for the great collaboration and incorporating our StreamDiffusionV2 pipeline into their cool Demo UI.

If you find this repository useful in your research, please consider giving a star ⭐ or a citation.

@article{feng2025streamdiffusionv2,
  title={StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation},
  author={Feng, Tianrui and Li, Zhi and Yang, Shuo and Xi, Haocheng and Li, Muyang and Li, Xiuyu and Zhang, Lvmin and Yang, Keting and Peng, Kelly and Han, Song and others},
  journal={arXiv preprint arXiv:2511.07399},
  year={2025}
}