fast-bagel

Fast-BAGEL (LGCC): Fast and Detail-Enhanced Image Editing

<p align="center"> <a href="https://zan.chat/fast-bagel"> <img src="https://img.shields.io/badge/Project-Website-0A66C2?logo=safari&logoColor=white" alt="LGCC Project Website"/> </a> <a href="https://arxiv.org/abs/2511.01894"> <img src="https://img.shields.io/badge/arXiv-2511.01894-b31b1b?logo=arxiv&logoColor=white" alt="LGCC Paper on arXiv"/> </a> <a href="https://huggingface.co/zanchat-ai/fast-bagel"> <img src="https://img.shields.io/badge/Model-Hugging%20Face-ffc107?logo=huggingface&logoColor=black" alt="LGCC (Fast-BAGEL) on Hugging Face"/> </a> <a href="https://github.com/ZanChat/fast-bagel"> <img src="https://img.shields.io/badge/GitHub-Repository-181717?logo=github&logoColor=white" alt="LGCC (Fast-BAGEL) GitHub Repository"/> </a> <a href="mailto:[email protected]"> <img src="https://img.shields.io/badge/Contact-Email%20Us-D14836?logo=gmail&logoColor=white" alt="Contact via Email"/> </a> </p>

This is the official Hugging Face repository for Fast-BAGEL, the model presented in the paper "LGCC: Enhancing Flow Matching Based Text-Guided Image Editing with Local Gaussian Coupling and Context Consistency."

LGCC (Fast-BAGEL) is a finetuned version of the ByteDance BAGEL model, designed to address three key challenges in flow matching-based image editing:

1. Slow Inference Speed: Existing models require many steps to produce high-quality results.
2. Detail Degradation: Fine details are often lost or blurred during the editing process.
3. Content Inconsistency: Edits can unintentionally alter parts of the image that should remain unchanged.

Our framework introduces two core innovations—Local Gaussian Noise Coupling (LGNC) and Content Consistency Loss (CCL)—to deliver faster, more detailed, and context-aware image edits.

Model Details

• Base Model: Finetuned from ByteDance-Seed/BAGEL-7B-MoT.
• Architecture: Flow matching-based model for text-guided image editing.
• Key Innovations:
  • Local Gaussian Noise Coupling (LGNC): A novel noise scheduling technique that preserves fine-grained spatial details by coupling noise to the local characteristics of the latent space. This is the primary driver of improved detail and faster convergence.
  • Content Consistency Loss (CCL): A new loss function applied during training that ensures semantic alignment between the text prompt and the image, preventing unintended alterations to background content.
• Training: The model was finetuned on a private dataset using a curriculum learning strategy to efficiently integrate our improvements into the pre-trained ByteDance BAGEL checkpoint.

How to Get Started

This model is designed to be used with our official project repository.

1. Clone the Repository

git clone https://github.com/ZanChat/fast-bagel.git
cd fast-bagel
# Install dependencies, e.g., pip install -r requirements.txt

2. Download Model Files

The model files in this repository will be automatically downloaded when you run the inference script. Ensure you have huggingface_hub installed.

3. Run Inference

Use the provided inference_ds.py script to perform text-guided image editing.

CUDA_VISIBLE_DEVICES=0 python inference_ds.py \
  --model_dir zanchat-ai/fast-bagel \
  --num_timesteps=50 \
  --time_shift=4

• --model_dir: The Hugging Face repository ID (zanchat-ai/fast-bagel).
• --num_timesteps: Number of inference steps. 50 steps are recommended for high-quality results (vs. 100 for the original BAGEL). Use fewer steps (e.g., 20-30) for faster previews.
• --time_shift: A parameter for the solver schedule. 4 is a good default.

Performance and Benchmarks

LGCC significantly outperforms baselines in both efficiency and quality.

Inference Speed-Up

Quality Improvement (on I2EBench)

Fig. 1: LGCC (Fast-BAGEL) achieves comparable or superior results in just 20-30 steps, while the original BAGEL requires 50 steps.

Limitations and Bias

• Inherited Limitations: As a finetuned version of ByteDance's BAGEL, this model may inherit any limitations or biases present in the original model and its training data.
• Content Domain: The model's performance is strongest on general, real-world images. It may struggle with highly abstract, technical, or niche visual concepts.
• Bias in Training Data: The model was trained on large-scale datasets which may contain societal, cultural, and gender biases. Generated or edited images may reflect these biases. We recommend users be mindful of this when using the model for downstream applications.
• Complex Edits: While significantly improved, the model may still face challenges with edits that require major topological changes, complex logical reasoning, or understanding of nuanced, abstract concepts.

Citation

If you find our work useful, please cite our paper:

@article{liu2025-fast-bagel,
  title   = {LGCC: Enhancing Flow Matching Based Text-Guided Image Editing with Local Gaussian Coupling and Context Consistency},
  author  = {Fangbing Liu and Pengfei Duan and Wen Li and Yi He},
  journal = {arXiv preprint arXiv:2511.01894},
  year    = {2025}
}

We also encourage you to cite the original ByteDance BAGEL paper, upon which our work is built:

@article{deng2025bagel,
  title   = {Emerging Properties in Unified Multimodal Pretraining},
  author  = {Deng, Chaorui and Zhu, Deyao and Li, Kunchang and Gou, Chenhui and Li, Feng and Wang, Zeyu and Zhong, Shu and Yu, Weihao and Nie, Xiaonan and Song, Ziang and Shi, Guang and Fan, Haoqi},
  journal = {arXiv preprint arXiv:2505.14683},
  year    = {2025}
}

Contact

For questions, collaborations, or issues, please contact us at [email protected] or open an issue in the GitHub repository.