---
title: "Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Model"
canonical_url: "https://www.modelscope.cn/papers/125511"
md_url: "https://www.modelscope.cn/papers/125511.md"
arxiv_id: 2503.07703
published: 2025-03-10
last_updated: 2025-03-10
authors:
  - "Lixue Gong"
  - "Xiaoxia Hou"
  - "Fanshi Li"
  - "Liang Li"
  - "Xiaochen Lian"
  - "Fei Liu"
  - "Liyang Liu"
  - "Wei Liu"
  - "Wei Lu"
  - "Yichun Shi"
  - "Shiqi Sun"
  - "Yu Tian"
  - "Zhi Tian"
  - "Peng Wang"
  - "Xun Wang"
  - "Ye Wang"
  - "Guofeng Wu"
  - "Jie Wu"
  - "Xin Xia"
  - "Xuefeng Xiao"
  - "Linjie Yang"
  - "Zhonghua Zhai"
  - "Xinyu Zhang"
  - "Qi Zhang"
  - "Yuwei Zhang"
  - "Shijia Zhao"
  - "Jianchao Yang"
  - "Weilin Huang"
model_name: "Seedream 2.0"
model_developer: ByteDance
domain:
  - "计算机视觉"
  - "图像生成"
  - "扩散模型"
  - "文本到图像"
  - "多模态"
type:
  - "计算机视觉"
  - "图像生成"
  - "扩散模型"
  - "文本到图像"
  - "多模态"
  - "Computer Vision and Pattern Recognition"
arxiv_url: "https://arxiv.org/abs/2503.07703"
pdf_url: "https://arxiv.org/pdf/2503.07703"
---

# Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Model

> Rapid advancement of diffusion models has catalyzed remarkable progress in the field of image generation. However, prevalent models such as Flux, SD3.5 and Midjourney, still grapple with issues like model bias, limited text rendering capabilities, and…

「Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Model」是 ModelScope 魔搭社区收录的论文，arXiv 2503.07703，作者为 Lixue Gong, Xiaoxia Hou, Fanshi Li et al.，发表于 2025-03-10，属于 计算机视觉、图像生成、扩散模型 领域。

- **ArXiv**: 2503.07703
- **Published**: 2025-03-10
- **Authors**: Lixue Gong, Xiaoxia Hou, Fanshi Li, Liang Li, Xiaochen Lian, Fei Liu, Liyang Liu, Wei Liu, Wei Lu, Yichun Shi, Shiqi Sun, Yu Tian, Zhi Tian, Peng Wang, Xun Wang, Ye Wang, Guofeng Wu, Jie Wu, Xin Xia, Xuefeng Xiao, Linjie Yang, Zhonghua Zhai, Xinyu Zhang, Qi Zhang, Yuwei Zhang, Shijia Zhao, Jianchao Yang, Weilin Huang
- **Model**: Seedream 2.0
- **Developer**: ByteDance
- **Domain**: 计算机视觉, 图像生成, 扩散模型, 文本到图像, 多模态
- **ArXiv URL**: https://arxiv.org/abs/2503.07703
- **PDF**: https://arxiv.org/pdf/2503.07703

Source: https://www.modelscope.cn/papers/125511

---

> Seedream 2.0：突破双语图像生成界限，精准捕捉中西文化精髓

## 摘要

本文针对现有图像生成模型中存在的问题，如模型偏差、文本渲染能力不足以及对中国文化理解的欠缺，提出了一种新型的中英双语图像生成基础模型——Seedream 2.0。该模型能够高效处理中英文提示，支持双语文本渲染和图像生成任务，并在多个方面表现出色。为解决上述问题，研究团队设计了一套能够持续整合新知识的数据架构，并开发了一个兼顾准确性和丰富性的强大标题系统。同时，集成了自主研发的多语言大语言模型（LLM）作为文本编码器，使其具备从高质量中文数据中直接学习的能力。此外，通过引入Glyph对齐的ByT5模型实现灵活的字符级文本渲染，并提出Scaled ROPE以适应未训练过的图像分辨率。在后训练阶段，通过多层次的监督微调（SFT）和基于人类反馈的强化学习（RLHF）迭代优化模型性能。

实验结果表明，Seedream 2.0在遵循提示、美学表现、文本渲染和结构正确性等方面均达到了行业领先水平。特别是在中文文本渲染方面，其表现显著优于社区内其他模型。经过多次RLHF优化，该模型的输出与人类偏好高度一致，这在ELO评分中得到了充分体现。此外，Seedream 2.0还可以轻松扩展为指令驱动的图像编辑模型，例如SeedEdit，具有强大的编辑能力。

总体而言，Seedream 2.0不仅解决了现有模型的局限性，还展示了在商业应用中的巨大潜力，有望成为提升生产力的有效工具。

## Abstract

Rapid advancement of diffusion models has catalyzed remarkable progress in the field of image generation. However, prevalent models such as Flux, SD3.5 and Midjourney, still grapple with issues like model bias, limited text rendering capabilities, and insufficient understanding of Chinese cultural nuances. To address these limitations, we present Seedream 2.0, a native Chinese-English bilingual image generation foundation model that excels across diverse dimensions, which adeptly manages text prompt in both Chinese and English, supporting bilingual image generation and text rendering. We develop a powerful data system that facilitates knowledge integration, and a caption system that balances the accuracy and richness for image description. Particularly, Seedream is integrated with a self-developed bilingual large language model as a text encoder, allowing it to learn native knowledge directly from massive data. This enable it to generate high-fidelity images with accurate cultural nuances and aesthetic expressions described in either Chinese or English. Beside, Glyph-Aligned ByT5 is applied for flexible character-level text rendering, while a Scaled ROPE generalizes well to untrained resolutions. Multi-phase post-training optimizations, including SFT and RLHF iterations, further improve the overall capability. Through extensive experimentation, we demonstrate that Seedream 2.0 achieves state-of-the-art performance across multiple aspects, including prompt-following, aesthetics, text rendering, and structural correctness. Furthermore, Seedream 2.0 has been optimized through multiple RLHF iterations to closely align its output with human preferences, as revealed by its outstanding ELO score. In addition, it can be readily adapted to an instruction-based image editing model, such as SeedEdit, with strong editing capability that balances instruction-following and image consistency.
