SkyworkUniPic

[🤗 UniPic checkpoint] [📖 Tech Report] [🏄 Web Demo]

Welcome to the Skywork-UniPic repository! This repository contains the model weights and implementation of our unified 1.5B-parameter autoregressive model that natively integrates image understanding, text-to-image generation, and image editing capabilities within a single architecture.

Evaluation

Performance Overview

GenEval

Model	Single	Two	Count	Color	Position	Attr	Overall
Diffusion Models
SDv2.1	0.98	0.51	0.44	0.85	0.07	0.17	0.50
SDXL	0.98	0.74	0.39	0.85	0.15	0.23	0.55
IF‑XL	0.97	0.74	0.66	0.81	0.13	0.35	0.61
LUMINA‑Next	0.92	0.46	0.48	0.70	0.09	0.13	0.46
SD3‑medium	0.99	0.94	0.72	0.89	0.33	0.60	0.74
FLUX.1‑dev	0.99	0.81	0.79	0.74	0.20	0.47	0.67
NOVA	0.99	0.91	0.62	0.85	0.33	0.56	0.71
Autoregressive Models
TokenFlow‑XL	0.95	0.60	0.41	0.81	0.16	0.24	0.55
Janus	0.97	0.68	0.30	0.84	0.46	0.42	0.61
Janus Pro	0.99	0.89	0.59	0.90	0.79	0.66	0.80
Emu3‑Gen	0.99	0.81	0.42	0.80	0.49	0.45	0.66
Show‑o	0.98	0.80	0.66	0.84	0.31	0.50	0.68
Unified Models
OmniGen	0.98	0.84	0.66	0.74	0.40	0.43	0.68
OmniGen2	1.00	0.95	0.64	0.88	0.55	0.76	0.80
OmniGen2†	0.99	0.96	0.74	0.98	0.71	0.75	0.86
MetaQuery‑XL†	-	-	-	-	-	-	0.80
BLIP3‑o† 4B	-	-	-	-	-	-	0.81
BLIP3‑o† 8B	-	-	-	-	-	-	0.84
BAGEL	0.99	0.94	0.81	0.88	0.64	0.63	0.82
BAGEL†	0.98	0.95	0.84	0.95	0.78	0.77	0.88
UniWorld‑V1	0.99	0.93	0.79	0.89	0.49	0.70	0.80
UniWorld‑V1†	0.98	0.93	0.81	0.89	0.74	0.71	0.84
Ovis‑U1	0.98	0.98	0.90	0.92	0.79	0.75	0.89
Proprietary Models
GPT‑4o	0.99	0.92	0.85	0.92	0.75	0.61	0.84
Skywork UniPic	0.98	0.92	0.74	0.91	0.89	0.72	0.86

DPG‑Bench

Model	Global	Entity	Attribute	Relation	Other	Overall
Diffusion Models
LUMINA‑Next	82.82	88.65	86.44	80.53	81.82	74.63
SDXL	83.27	82.43	80.91	86.76	80.41	74.65
PlayGroundv2.5	83.06	82.59	81.20	84.08	83.50	75.47
Hunyuan‑DiT	84.59	80.59	88.01	74.36	86.41	78.87
PixArt‑Σ	86.89	82.89	88.94	86.59	87.68	80.54
DALLE3	90.97	89.61	88.39	90.58	89.83	83.50
SD3‑medium	87.90	91.01	88.83	80.70	88.68	84.08
FLUX.1‑dev	82.10	89.50	88.70	91.10	89.40	84.00
Autoregressive Models
Show‑o	79.33	75.44	78.02	84.45	60.80	67.27
EMU3	85.21	86.68	86.84	90.22	83.15	80.60
TokenFlow‑XL	78.72	79.22	81.29	85.22	71.20	73.38
Janus	82.33	87.38	87.70	85.46	86.41	79.68
Janus Pro	86.90	88.90	89.40	89.32	89.48	84.19
BLIP3‑o 4B	-	-	-	-	-	79.36
BLIP3‑o 8B	-	-	-	-	-	81.60
Unified Models
OmniGen	87.90	88.97	88.47	87.95	83.56	81.16
OmniGen2	88.81	88.83	90.18	89.37	90.27	83.57
BAGEL	88.94	90.37	91.29	90.82	88.67	85.07
UniWorld‑V1	83.64	88.39	88.44	89.27	87.22	81.38
Ovis‑U1	82.37	90.08	88.68	93.35	85.20	83.72
Skywork UniPic	89.65	87.78	90.84	91.89	91.95	85.50

GEdit‑Bench‑EN

Model	SC ↑	PQ ↑	Overall ↑
Proprietary Models
Gemini‑2.0‑flash	6.73	6.61	6.32
GPT‑4o	7.85	7.62	7.53
Specialized Editing Models
Instruct‑Pix2Pix	3.58	5.49	3.68
MagicBrush	4.68	5.66	4.52
AnyEdit	3.18	5.82	3.21
ICEdit	5.11	6.85	4.84
Step1X‑Edit	7.09	6.76	6.70
Unified Models
OmniGen	5.96	5.89	5.06
OmniGen2	7.16	6.77	6.41
BAGEL	7.36	6.83	6.52
UniWorld‑V1	4.93	7.43	4.85
Ovis‑U1	-	-	6.42
Skywork UniPic	6.72	6.18	5.83

Usage

📦 Required Packages

Create virtual environment and install dependencies with pip:

conda create -n unipic python==3.10.14
conda activate unipic
pip install -r requirements.txt

📥 Checkpoints

Download the model checkpoints from [🤗 SkyworkUniPic], It is recommended to use the following command to download the checkpoints

# pip install -U "huggingface_hub[cli]"
huggingface-cli download Skywork/Skywork-UniPic-1.5B  --local-dir checkpoint --repo-type model

🖼️ Text-to-image Generation

You can generate images from text prompts using the following command:

# only support 1024 image_size
export PYTHONPATH=./:$PYTHONPATH
python scripts/text2image.py configs/models/qwen2_5_1_5b_kl16_mar_h.py \
         --checkpoint checkpoint/pytorch_model.bin  --image_size 1024 \
         --prompt 'A glossy-coated golden retriever stands on the park lawn beside a life-sized penguin statue.'  --output output.jpg

To generate a list of images based on prompts in a json file.

export PYTHONPATH=./:$PYTHONPATH
accelerate launch scripts/batch_text2image.py configs/models/qwen2_5_1_5b_kl16_mar_h.py \
       --checkpoint checkpoint/pytorch_model.bin  --image_size 1024 \
       --data data/batch_t2i.json --output output --batch_size 2 --grid_size 2

The json file should look like:

[
  {
   "prompt": "A glossy-coated golden retriever stands on the park lawn beside a life-sized penguin statue."
  },
  {
   "prompt": "Digital portrait of a girl with rainbow hair."
  }
]

✏️ Image Editing

The image editing feature within this unified model is an exploratory module at the forefront of research. It currently faces challenges in edit precision, control granularity, and output consistency, and is not yet production-ready.

# only support 1024 image_size
export PYTHONPATH=./:$PYTHONPATH
python scripts/image_edit.py configs/models/qwen2_5_1_5b_kl16_mar_h.py \
         --checkpoint checkpoint/pytorch_model.bin  --image_size 1024 \
         --image data/sample.png --prompt "Replace the stars with the candle." \
         --output output.jpg 

📖 Image-to-text Generation

export PYTHONPATH=./:$PYTHONPATH
python scripts/image2text.py configs/models/qwen2_5_1_5b_kl16_mar_h.py \
         --checkpoint checkpoint/pytorch_model.bin  --image_size 1024 \
         --image data/sample.png --prompt "Describe the image in detail."

📜 License

This project is licensed under MIT License.

Citation

If you use Skywork-UniPic in your research, please cite:

@misc{skywork2025unipic,
  title={Skywork-UniPic: Unified Autoregressive Modeling for Visual Understanding and Generation},
  author={Skywork Multimodality Team},
  year={2025},
  publisher={GitHub},
  journal={GitHub repository},
  howpublished={\url{https://github.com/SkyworkAI/UniPic}},
  note={MIT License}
}