1. Tutorial Introduction

LongCat-Image is an open-source image generation and editing model released by Meituan's LongCat team in December 2025. Designed for bilingual (Chinese and English) scenarios, it boasts exceptional text-to-image generation and text rendering capabilities. With only 6 bytes of parameters, this model demonstrates efficiency and performance far exceeding similar open-source models, achieving high-quality, realistic visual generation results in multiple benchmark tests, and reaching industry-leading levels in the accuracy and coverage of Chinese text rendering. Furthermore, LongCat-Image provides advanced image editing capabilities and a comprehensive open-source toolchain, enabling developers to deploy, research, and further develop the model with lower barriers to entry, bringing efficient, realistic, and high-quality image output to the open-source ecosystem. Related research papers are available. LongCat-Image Technical Report .

This tutorial uses a single RTX 5090 graphics card as the default resource.

2. Project Examples

3. Operation steps

1. After starting the container, click the API address to enter the Web interface

2. After entering the webpage, you can enter text and generate an image.

If "Bad Gateway" is displayed, it means the model is initializing. Since the model is large, please wait about 3-4 minutes and refresh the page.

How to use

Parameter Description

• Custom LoRA (optional)
  • Custom LoRA: Enter the URL or path for LoRA weights to load LoRA models with additional styles or capabilities.
  • LoRA ScaleLoRA intensity (range 0-2)
• Output resolution
  • Width: Width of the generated image (64~2048, you can enter it yourself or drag the slider)
  • Height: Height of the generated image (64~2048, can be entered manually or by dragging the slider)
• Random seed settings
  • Seed: Controlling the randomness of generated images
    • -1 or check "Randomize" to indicate a random seed each time.
    • Entering a fixed number will reproduce the same image.
  • Randomize seedWhen checked, a different seed will be used for each generation.
• Inference parameters
  • Inference Steps: Affects the generation quality and speed (range 1-100, the higher the value, the higher the image quality usually is but the longer it takes).
  • Guidance ScaleControls the degree of influence of "text hints" on images (range 1-20).
    • The higher the value, the more closely it matches the prompt word.
    • Lower values indicate more freedom and greater randomness.

Citation Information

@article{LongCat-Image,
      title={LongCat-Image Technical Report},
      author={Meituan LongCat Team and  Hanghang Ma and Haoxian Tan and Jiale Huang and Junqiang Wu and Jun-Yan He and Lishuai Gao and Songlin Xiao and Xiaoming Wei and Xiaoqi Ma and Xunliang Cai and Yayong Guan and Jie Hu},
	    journal={arXiv preprint arXiv:2512.07584},
      year={2025}
}