.svg){kind=icon}
Haoge Deng1,4*, Ting Pan2,4*, Fan Zhang4*, Yang Liu3,4*, Zhuoyan Luo4, Yufeng Cui4, Wenxuan Wang4
Chunhua Shen3, Shiguang Shan2, Zhaoxiang Zhang1†, Xinlong Wang4†
CASIA1, CASICT2, ZJU3, BAAI4
* Equal Contribution, † Corresponding Author
We present URSA (Uniform discRete diffuSion with metric pAth), a simple yet powerful framework that bridges the gap with continuous approaches. URSA formulates the video generation task as an iterative global refinement of discrete spatiotemporal tokens and scales efficiently to long video generation, requiring fewer inference steps. URSA enables multi-task video generation with asynchronous timestep scheduling strategy in one unified model.
[Oct 2025]🎉 URSA is part of Emu3.5 as DiDA (Discrete Diffusion Adaptation)![Oct 2025]Released TI2V 🤗 Demo.[Oct 2025]Released Paper & Project Page & Evaluation Guide.
- 🥇 Novel Approach: Uniform Discrete Diffusion with Metric Path.
- 🥈 SOTA Performance: High efficiency with state-of-the-art T2I/T2V/I2V results.
- 🥉 Unified Modeling: Multi-task capabilities in a single unified model.
| Model | Resolution | Data | Weight | GenEval | DPGBench |
|---|---|---|---|---|---|
| URSA-0.6B-IBQ1024 | 1024x1024 | 30M | 🤗 HF | 🤖 ModelScope | 0.79 | 85.6 |
| URSA-1.7B-IBQ1024 | 1024x1024 | 30M | 🤗 HF | 🤖 ModelScope | 0.80 | 86.0 |
| Model | Resolution | Data | Weight | VBench-T2V | VBench-I2V |
|---|---|---|---|---|---|
| URSA-0.6B-FSQ320 | 49x512x320 | 24M | 🤗 HF | 🤖 ModelScope | 81.4 | 86.0 |
| URSA-1.7B-FSQ320 | 49x512x320 | 24M | 🤗 HF | 🤖 ModelScope | 82.4 | 86.2 |
Clone this repository to local disk and install:
pip install diffusers transformers>=4.57.1 accelerate imageio imageio-ffmpeg omegaconf wandb
git clone https://github.com/baaivision/URSA.git
cd URSA && pip install .import torch
from diffnext.pipelines import URSAPipeline
model_id, height, width = "BAAI/URSA-1.7B-IBQ1024", 1024, 1024
model_args = {"torch_dtype": torch.float16, "trust_remote_code": True}
pipe = URSAPipeline.from_pretrained(model_id, **model_args)
pipe = pipe.to(torch.device("cuda"))
prompt = "The bear, calm and still, gazes upward as if lost in contemplation of the cosmos."
negative_prompt = "worst quality, low quality, inconsistent motion, static, still, blurry, jittery, distorted, ugly"
image = pipe(**locals()).frames[0]
image.save("ursa.jpg")import os, torch, numpy
from diffnext.pipelines import URSAPipeline
from diffnext.utils import export_to_video
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
model_id, height, width = "BAAI/URSA-1.7B-FSQ320", 320, 512
model_args = {"torch_dtype": torch.float16, "trust_remote_code": True}
pipe = URSAPipeline.from_pretrained(model_id, **model_args)
pipe = pipe.to(torch.device("cuda"))
text_prompt = "a lone grizzly bear walks through a misty forest at dawn, sunlight catching its fur."
negative_prompt = "worst quality, low quality, inconsistent motion, static, still, blurry, jittery, distorted, ugly"
# Text-to-Image
prompt = text_prompt
num_frames, num_inference_steps = 1, 25
image = pipe(**locals()).frames[0]
image.save("ursa.jpg")
# Image-to-Video
prompt = f"motion=9.0, {text_prompt}"
num_frames, num_inference_steps = 49, 50
video = pipe(**locals()).frames[0]
export_to_video(video, "ursa_1+48f.mp4", fps=12)
# Text-to-Video
image, video = None, None
prompt = f"motion=9.0, {text_prompt}"
num_frames, num_inference_steps = 49, 50
video = pipe(**locals()).frames[0]
export_to_video(video, "ursa_49f.mp4", fps=12)
# Video-to-Video
prompt = f"motion=5.0, {text_prompt}"
num_frames, num_inference_steps = 49, 50
num_cond_frames, cond_noise_scale = 13, 0.1
for i in range(12):
video, start_video = video[-num_cond_frames:], video
video = pipe(**locals()).frames[0]
video = numpy.concatenate([start_video, video[num_cond_frames:]])
export_to_video(video, "ursa_{}f.mp4".format(video.shape[0]), fps=12)# Text-to-Image (T2I)
python scripts/app_ursa_t2i.py --model "BAAI/URSA-1.7B-IBQ1024" --device 0
# Text-to-Image-to-Video (TI2V)
python scripts/app_ursa_ti2v.py --model "BAAI/URSA-1.7B-FSQ320" --device 0If you find this repository useful, please consider giving a star ⭐ and citation 🦖:
@article{deng2025ursa,
title={Uniform Discrete Diffusion with Metric Path for Video Generation},
author={Deng, Haoge and Pan, Ting and Zhang, Fan and Liu, Yang and Luo, Zhuoyan and Cui, Yufeng and Shen, Chunhua and Shan, Shiguang and Zhang, Zhaoxiang and Wang, Xinlong},
journal={arXiv preprint arXiv:2510.24717},
year={2025}
}
@article{deng2024nova,
title={Autoregressive Video Generation without Vector Quantization},
author={Deng, Haoge and Pan, Ting and Diao, Haiwen and Luo, Zhengxiong and Cui, Yufeng and Lu, Huchuan and Shan, Shiguang and Qi, Yonggang and Wang, Xinlong},
journal={arXiv preprint arXiv:2412.14169},
year={2024}
}
We thank the repositories:
- NOVA. ✨NOVA is the predecessor of 🐻URSA.
- FlowMatching. This codebase systemically provides CFM and DFM implementations.
- FUDOKI. This codebase provides a naive multimodal DFM implementation.
- CodeWithGPU. CodeWithGPU library is the core of our data loading pipeline.
Code and models are licensed under Apache License 2.0.