伦敦——今日，人工智能研究机构Stability AI宣布推出Stable Diffusion 3的研究论文，该模型在文生图领域取得了显著的技术进步。这一最新版本在理解和生成与文本提示相符的高质量图像方面，据称超越了当前的行业标杆，如DALL·E 3、Midjourney v6和Ideogram v1。
Stable Diffusion 3的核心创新在于其多模态扩散Transformer（MMDiT）架构，这一架构使用独立的权重集分别处理图像和语言表示。这一改进显著提升了模型的文本理解能力和对拼写的敏感度，从而在生成图像时能更准确地捕捉和呈现用户的文字描述。
根据Stability AI提供的信息，Stable Diffusion 3在人类偏好评估中表现出色，尤其是在排版和遵循提示方面。这一成就表明，该模型在创造与文本内容更加贴切且艺术性更强的图像方面，已达到新的高度。
这一技术突破对于推动人工智能在艺术、设计和创意产业的应用具有重大意义，可能开启一个由人工智能辅助的全新创作时代。随着Stable Diffusion 3的发布，Stability AI再次证明了其在先进人工智能研究领域的领先地位，为未来的文生图技术设定了新的标准。
英语如下：
**News Title:** “Stability AI Launches Stable Diffusion 3: A Breakthrough in Text-to-Image Generation, Outperforming Rivals like DALL·E 3”
**Keywords:** Stability AI, Stable Diffusion 3, Text-to-Image Model
**News Content:**
**Title:** Stability AI Unveils Stable Diffusion 3 Text-to-Image Model, Advancing Quality in Text-to-Image Generation
**London** — Artificial intelligence research firm Stability AI has announced the release of its Stable Diffusion 3 research paper, marking a significant advancement in the field of text-to-image models. The latest iteration is said to surpass current industry leaders, such as DALL·E 3, Midjourney v6, and Ideogram v1, in generating high-quality images that align with textual prompts.
At the heart of Stable Diffusion 3 is its Multi-Modal Diffusion Transformer (MMDiT) architecture, which employs separate weight sets for handling image and language representations. This innovation enhances the model’s text comprehension and sensitivity to spelling, enabling it to more accurately capture and depict user descriptions when generating images.
According to information provided by Stability AI, Stable Diffusion 3 excels in human preference assessments, particularly in layout and adherence to prompts. This achievement signals that the model has reached new heights in creating images that are more contextually accurate and artistically refined based on textual content.
This technological breakthrough holds significant implications for the application of AI in the arts, design, and creative industries, potentially ushering in a new era of AI-assisted creation. With the launch of Stable Diffusion 3, Stability AI further solidifies its position at the forefront of advanced AI research, setting new standards for the future of text-to-image technology.
【来源】https://stability.ai/news/stable-diffusion-3-research-paper
Views: 1