AIDC-AI / Ovis-U1-3B

Ovis-U1

Building on the foundation of the Ovis series, Ovis-U1 is a 3-billion-parameter unified model that seamlessly integrates multimodal understanding , text-to-image generation , and image editing within a single powerful framework.

The overall architecture of Ovis-U1 (cf. Fig.2 in our report).

📦 Installation

Ovis-U1 has been tested with Python 3.10, Torch 2.4.0, Transformers 4.51.3, and DeepSpeed 0.15.4. For a comprehensive list of package dependencies, please consult the requirements.txt file.

git clone [email protected]:AIDC-AI/Ovis-U1.git
conda create -n ovis-u1 python=3.10 -y
conda activate ovis-u1
cd Ovis-U1
pip install -r requirements.txt
pip install -e .

🛠️ Inference

For multimodal understanding, please run

python test_img_to_txt.py

For text-to-image, please run

python test_txt_to_img.py \
    --height 1024 \
    --width 1024  \
    --steps 50 \
    --seed 42 \
    --txt_cfg 5  

For image editing, please run

python test_img_edit.py \
    --steps 50 \
    --img_cfg 1.5 \
    --txt_cfg 6  

📊 Performance

OpenCompass Multi-modal Academic Benchmarks

GenEval

📝 denotes using the rewritten prompts

DPG-Bench

ImgEdit-Bench

GEdit-Bench-EN

📚 Citation

If you find Ovis-U1 useful, please cite our paper:

@inproceedings{wang2025ovisu1,
title={Ovis-U1 Technical Report},
author={Ovis Team},
year={2025}
}

🙏 Acknowledgments

The code is built upon Ovis and FLUX .

📄 License

The project is released under Apache License 2.0 ( http://www.apache.org/licenses/LICENSE-2.0 , SPDX-License-identifier: Apache-2.0).

🚨 Disclaimer

We used compliance checking algorithms during the training process, to ensure the compliance of the trained model to the best of our ability. Due to complex data and the diversity of language model usage scenarios, we cannot guarantee that the model is completely free of copyright issues or improper content. If you believe anything infringes on your rights or generates improper content, please contact us, and we will promptly address the matter.