UCSC-VLAA/gpt-image-edit-training

GPT-Image-Edit — Model Release

Paper: GPT-IMAGE-EDIT-1.5M: A Million-Scale, GPT-Generated Image Dataset
Code: wyhlovecpp/GPT-Image-Edit
Data: UCSC-VLAA/GPT-Image-Edit-1.5M
Project Page: https://ucsc-vlaa.github.io/GPT-Image-Edit

About GPT-Image-Edit

Recent advancements in large multimodal models like GPT-4o have set a new standard for high-fidelity, instruction-guided image editing. To bridge the gap for open-source research, we introduce GPT-IMAGE-EDIT-1.5M, a publicly available, large-scale image-editing corpus containing more than 1.5 million high-quality triplets (instruction, source image, edited image). We systematically construct this dataset by leveraging the versatile capabilities of GPT-4o to unify and refine three popular image-editing datasets: OmniEdit, HQ-Edit, and UltraEdit. Specifically, our methodology involves 1) regenerating output images to enhance visual quality and instruction alignment, and 2) selectively rewriting prompts to improve semantic clarity. To validate the efficacy of our dataset, we fine-tune advanced open-source models on GPT-IMAGE-EDIT-1.5M. The empirical results are exciting, e.g., the fine-tuned FluxKontext achieves highly competitive performance across a comprehensive suite of benchmarks, showing stronger instruction following and higher perceptual quality while maintaining identity. These scores markedly exceed all previously published open-source methods and substantially narrow the gap to leading proprietary models. We hope the full release of GPT-IMAGE-EDIT-1.5M can help to catalyze further open research in instruction-guided image editing.

📣 News

• [2025.07.27] 🤗 We release GPT-Image-Edit, a state-of-the-art image editing model with 1.5M high-quality editing samples. All data, models, training code and evaluation code are open-sourced. Our code is based on UniWorld-V1, Thanks to the authors of UniWorld-V1. Checking our report for more details. Welcome to watch 👀 this repository for the latest updates.

📦 Repository Structure

Folder	Description
final_checkpoint/	Final model weights for inference/sampling
stage1_checkpoint/	Qwen-VL → Flux connector (for Stage-2 fine-tuning)
training_json/	JSON files for training (compatible with repo dataloader)

🔥 Quick Start

1.Set up environment

git clone https://github.com/wyhlovecpp/GPT-Image-Edit.git
cd GPT-Image-Edit
conda create -n univa python=3.10 -y
conda activate univa
pip install -r requirements.txt
pip install flash_attn --no-build-isolation

2.Download pretrained checkpoint

huggingface-cli download --resume-download UCSC-VLAA/gpt-image-edit-training --local-dir ${MODEL_PATH}
huggingface-cli download --resume-download black-forest-labs/FLUX.1-Kontext-dev --local-dir ${FLUX_PATH}

3.Run with CLI

MODEL_PATH="path/to/model/final_checkpoint"
FLUX_PATH="path/to/flux"
CUDA_VISIBLE_DEVICES=0 python -m univa.serve.cli \
    --model_path ${MODEL_PATH} \
    --flux_path ${FLUX_PATH}

4.Run with Gradio Web Interface Highly recommend trying out our web demo by the following command.

python app.py --model_path ${MODEL_PATH} --flux_path ${FLUX_PATH}

5.Run with Gradio Serve

CUDA_VISIBLE_DEVICES=0 python -m univa.serve.gradio_web_server \
    --model_path ${MODEL_PATH} \
    --flux_path ${FLUX_PATH}

🗝️ Training

Data preparation

Download the data from UCSC-VLAA/GPT-Image-Edit-1.5M. The dataset consists of two parts: source images and annotation metadata. The json files is under UCSC-VLAA/gpt-image-edit-training/training_json

Prepare a data.txt file in the following format:

1. The first column is the root path to the image.

2. The second column is the corresponding annotation JSON file.

3. The third column indicates whether to enable the region-weighting strategy. We use False in our training setting.

We have prepared a data.txt file about gpt-edit for your reference.

`data.txt` for gpt-edit

data/gpt-edit/hqedit/edit,training_json/hqedit_gpt_edit.json,false
data/gpt-edit/hqedit/generate,training_json/hqedit_gpt_generate.json,false
data/gpt-edit/omniedit,training_json/omniedit_gpt.json,false
data/gpt-edit/omniedit,training_json/omniedit_gpt_rewrite.json,false
data/gpt-edit/omniedit/complex-edit,training_json/complexedit_gpt.json,false
data/gpt-edit/ultraedit,training_json/ultraedit_gpt.json,false

Data details

Image Editing

• UCSC-VLAA/GPT-Image-Edit-1.5M [5T storage usage.]

Training

Prepare pretrained weights

Download black-forest-labs/FLUX.1-Kontext-dev to $FLUX_PATH. Download Qwen/Qwen2.5-VL-7B-Instruct to $QWENVL_PATH. We also support other sizes of Qwen2.5-VL.

SAVE_PATH="path/to/save/UniWorld-Qwen2.5-VL-7B-Instruct-FLUX.1-dev-fp32"
python scripts/make_univa_qwen2p5vl_weight.py \
    --origin_flux_ckpt_path $FLUX_PATH \
    --origin_qwenvl_ckpt_path $QWENVL_PATH \
    --save_path ${SAVE_PATH}

Stage 1

You need to specify pretrained_lvlm_name_or_path to ${SAVE_PATH} in flux_qwen2p5vl_7b_vlm_stage1_512.yaml.

We recommend using pretrained weight from LanguageBind/UniWorld-V1/stage1.

# stage 1
# if use prodigy, pip install prodigy
bash scripts/denoiser/flux_qwen2p5vl_7b_vlm_stage1_512.sh

Stage 2

You need to specify pretrained_mlp2_path, which is trained by stage 1 or use the pretrained weight from LanguageBind/UniWorld-V1/stage1.

For training with 512×512 scale images (batch size 1), it consume about 78G in 1 node (8 GPUs).

Setting ema_pretrained_lvlm_name_or_path: null can saving memory if you want to train the higher resolution (e.g, 1024×1024 scale) or larger batch size. Using more nodes also can save memory because we use zero2 for main model in stage 2.

# stage 2
bash scripts/denoiser/flux_qwen2p5vl_7b_vlm_stage2_1024.sh

⚡️ Evaluation

Image Editing

ImgEdit

cd univa/eval/imgedit
# follow the instruction in univa/eval/imgedit/README.md

GEdit

cd univa/eval/gdit
# follow the instruction in univa/eval/gdit/README.md

complex-edit

cd univa/eval/complex-edit
# follow the instruction in univa/eval/complex-edit/README.md

OmniContext

cd univa/eval/omnicontext
# follow the instruction in univa/eval/omnicontext/README.md

📊 Benchmarks

GEdit-EN-full

Model	BG Change	Color Alt.	Mat. Mod.	Motion	Portrait	Style	Add	Remove	Replace	Text	Tone	Avg
Open-Sourced Models
AnyEdit	4.31	4.25	2.64	0.67	1.90	1.95	3.72	3.75	3.23	0.77	4.21	2.85
MagicBrush	6.17	5.41	4.75	1.55	2.90	4.10	5.53	4.13	5.10	1.33	5.07	4.19
Instruct-Pix2Pix	3.94	5.40	3.52	1.27	2.62	4.39	3.07	1.50	3.48	1.13	5.10	3.22
OmniGen	5.23	5.93	5.44	3.12	3.17	4.88	6.33	6.35	5.34	4.31	4.96	5.01
Step1X-Edit	7.03	6.26	6.46	3.66	5.23	7.24	7.17	6.42	7.39	7.40	6.62	6.44
Bagel	7.44	6.99	6.26	5.09	4.82	6.04	7.94	7.37	7.31	7.16	6.17	6.60
Bagel-thinking	7.22	7.24	6.69	7.12	6.03	6.17	7.93	7.44	7.45	3.61	6.36	6.66
Ovis-U1	7.49	6.88	6.21	4.79	5.98	6.46	7.49	7.25	7.27	4.48	6.31	6.42
OmniGen2	-	-	-	-	-	-	-	-	-	-	-	6.42
Step1X-Edit (v1.1)	7.45	7.38	6.95	4.73	4.70	7.11	8.20	7.59	7.80	7.91	6.85	6.97
FluxKontext dev	7.06	7.03	5.52	5.62	4.68	5.55	6.95	6.76	6.13	6.10	7.48	6.26
Proprietary Models
Gemini	7.11	7.14	6.47	5.67	3.99	4.95	8.12	6.89	7.41	6.85	7.01	6.51
Doubao	8.07	7.36	7.20	5.38	6.28	7.20	8.05	7.71	7.87	4.01	7.67	6.98
GPT-4o	6.96	6.85	7.10	5.41	6.74	7.44	7.51	8.73	8.55	8.45	8.69	7.49
Ours	7.80	7.54	7.12	7.75	7.09	6.74	8.04	7.95	7.17	5.45	6.95	7.24

Complex-Edit

Method	IF	IP	PQ	Overall
AnyEdit	1.60	8.15	7.25	5.67
UltraEdit	6.56	5.93	7.29	6.59
OmniGen	6.25	6.42	7.54	6.74
FluxKontext Dev	8.56	8.39	8.51	8.49
Imagen3	7.56	6.55	7.67	7.26
SeedEdit	8.49	6.91	8.74	8.04
GPT-4o	9.29	7.51	9.47	8.76
Ours	8.99	8.41	8.93	8.78

ImgEdit-Full

Model	Add	Adjust	Extract	Replace	Remove	Background	Style	Hybrid	Action	Overall
MagicBrush	2.84	1.58	1.51	1.97	1.58	1.75	2.38	1.62	1.22	1.90
Instruct-Pix2Pix	2.45	1.83	1.44	2.01	1.50	1.44	3.55	1.20	1.46	1.88
AnyEdit	3.18	2.95	1.88	2.47	2.23	2.24	2.85	1.56	2.65	2.45
UltraEdit	3.44	2.81	2.13	2.96	1.45	2.83	3.76	1.91	2.98	2.70
OmniGen	3.47	3.04	1.71	2.94	2.43	3.21	4.19	2.24	3.38	2.96
Step1X-Edit	3.88	3.14	1.76	3.40	2.41	3.16	4.63	2.64	2.52	3.06
ICEdit	3.58	3.39	1.73	3.15	2.93	3.08	3.84	2.04	3.68	3.05
BAGEL	3.56	3.31	1.70	3.30	2.62	3.24	4.49	2.38	4.17	3.20
UniWorld-V1	3.82	3.64	2.27	3.47	3.24	2.99	4.21	2.96	2.74	3.26
OmniGen2	3.57	3.06	1.77	3.74	3.20	3.57	4.81	2.52	4.68	3.44
Ovis-U1	4.13	3.62	2.98	4.45	4.06	4.22	4.69	3.45	4.61	4.00
FluxKontext dev	3.76	3.45	2.15	3.98	2.94	3.78	4.38	2.96	4.26	3.52
GPT-4o	4.61	4.33	2.90	4.35	3.66	4.57	4.93	3.96	4.89	4.20
Ours	4.07	3.79	2.04	4.13	3.89	3.90	4.84	3.04	4.52	3.80

👍 Acknowledgement and Related Work

• UniWorld-V1: UniWorld-V1 is a unified framework for understanding, generation, and editing.
• ImgEdit: ImgEdit is a large-scale, high-quality image-editing dataset comprising 1.2 million carefully curated edit pairs and a comprehensive benchmark for image editing.
• Complex-edit: Complex-edit is benchmark for complex image editing.
• Qwen2.5-VL: The new flagship vision-language model of Qwen.
• FLUX.1-Kontext-dev: A state-of-the-art image editing model.
• Step1X-Edit: A state-of-the-art image editing model and a comprehensive benchmark for image editing.
• OmniGen2: A state-of-the-art image editing model and a comprehensive benchmark for image editing.

🔒 License

• See LICENSE for details. The FLUX Kontext weights fall under the FLUX.1 Kontext [dev] Non-Commercial License.

✏️ Citing

@misc{wang2025gptimageedit15mmillionscalegptgeneratedimage,
      title={GPT-IMAGE-EDIT-1.5M: A Million-Scale, GPT-Generated Image Dataset},
      author={Yuhan Wang and Siwei Yang and Bingchen Zhao and Letian Zhang and Qing Liu and Yuyin Zhou and Cihang Xie},
      year={2025},
      eprint={2507.21033},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2507.21033},
}