【伦敦讯】今日，人工智能研究公司Stability AI在其官方网站上发布了备受瞩目的Stable Diffusion 3文生图模型的研究论文，标志着文本到图像生成技术的又一重大突破。Stable Diffusion 3模型在图像生成的质量和对提示的精准响应上，据评估超越了当前业界的领先系统，包括DALL·E 3、Midjourney v6和Ideogram v1。
据论文所述，Stable Diffusion 3的核心创新在于其全新的多模态扩散Transformer（MMDiT）架构。这一架构允许模型使用独立的权重集来处理图像和语言的表示，从而提升了模型在理解文本和拼写准确性上的能力。这一改进对于确保生成图像与输入文本的精确对应至关重要，进一步优化了用户体验。
Stability AI的研究表明，Stable Diffusion 3在人类偏好评估中表现出色，特别是在图像的排版布局和对用户输入提示的忠实度上。这一成果预示着未来人工智能在艺术创作、设计辅助以及视觉内容生成等领域将有更广泛的应用。
Stability AI的这一创新发布，无疑为人工智能与艺术的交叉领域开辟了新的可能，同时也对文本生成图像技术的未来发展提出了更高的标准。随着技术的不断进步，我们有理由期待AI在创造性的任务中展现出更加逼近人类智慧的表现。
英语如下：
**News Title:** “Stability AI Launches Stable Diffusion 3: A New Benchmark in Text-to-Image Generation超越DALL·E 3”
**Keywords:** Stability AI, Stable Diffusion 3, Text-to-Image Model
**News Content:**
**[London]** Today, artificial intelligence research firm Stability AI unveiled the much-anticipated Stable Diffusion 3 text-to-image model on its official website, marking a significant breakthrough in the field. The Stable Diffusion 3 model is assessed to surpass current industry leaders, including DALL·E 3, Midjourney v6, and Ideogram v1, in both image quality and precise responsiveness to prompts.
According to the research paper, the core innovation of Stable Diffusion 3 lies in its novel Multi-modal Diffusion Transformer (MMDiT) architecture. This architecture enables the model to handle image and language representations with separate weight sets, enhancing its capabilities in understanding text and improving spelling accuracy. This improvement is crucial for ensuring precise correspondence between generated images and input texts, thereby optimizing user experience.
Stability AI’s research demonstrates that Stable Diffusion 3 excels in human preference assessments, particularly in layout and composition of images, as well as faithfulness to user input prompts. This achievement foreshadows broader applications of AI in artistic creation, design assistance, and visual content generation.
Stability AI’s innovative release opens new possibilities in the intersection of AI and art, simultaneously setting higher standards for the future development of text-to-image generation technology. As technology progresses, there is reason to anticipate AI demonstrating ever-closer approximations to human creativity in tasks that require imagination.
【来源】https://stability.ai/news/stable-diffusion-3-research-paper
Views: 2