【麻省理工与Adobe联手推出DMD AI技术，图像生成速度大幅提升】
近日，国际知名学术机构麻省理工大学与全球领先的创意软件公司Adobe共同宣布，他们成功研发出一种名为DMD（Distribution Matching Distillation）的新型人工智能技术，极大地提升了图像生成的速度。这项技术针对主流文生图模型渲染速度缓慢的问题，提出了一种创新的解决方案，将多步扩散模型简化为一步图像生成过程。
DMD全称为“分布匹配蒸馏”，它在保证图像质量的同时，显著提高了生成速度。据研究人员透露，该技术能够在现代GPU硬件上实现每秒生成20幅图像的高效性能，这一速度相较于现有的图像生成技术有了显著提升，为人工智能在图像创作领域的应用开辟了新的可能。
这一突破性的进展对于新闻报道、影视制作、游戏设计以及艺术创作等领域具有重大意义，它将为内容创作者提供更快捷、更高效的工具，进一步推动数字内容生产的革新。麻省理工和Adobe的此项合作再次展示了科技与艺术结合的力量，预示着未来图像生成技术将更加智能化和高效化。
来源：IT之家
英语如下：
**News Title:** “MIT and Adobe Collaborate on DMD AI: Generating 20 Images per Second, Breaking Speed Records for Image Creation”
**Keywords:** MIT, Adobe, DMD AI
**News Content:**
**MIT and Adobe Join Forces to Launch DMD AI Technology, Significantly Boosting Image Generation Speed**
Recently, the renowned academic institution Massachusetts Institute of Technology (MIT) and global leader in creative software Adobe jointly announced the successful development of a groundbreaking artificial intelligence technology called DMD (Distribution Matching Distillation), which dramatically accelerates the speed of image generation. This innovation addresses the issue of slow rendering in mainstream text-to-image models by streamlining the multi-step diffusion process into a single-step image generation.
DMD, standing for “Distribution Matching Distillation,” maintains image quality while notably enhancing speed. According to researchers, the technology achieves an impressive performance of generating 20 images per second on modern GPU hardware, a significant improvement over existing image generation techniques. This breakthrough opens up new possibilities for AI applications in image creation.
This advancement holds great significance for sectors such as news reporting, film production, game design, and artistic creation, as it offers content creators faster and more efficient tools, further revolutionizing digital content production. The collaboration between MIT and Adobe underscores the power of combining technology with art, foreshadowing a future where image generation technologies will become even more intelligent and efficient.
**Source:** IT之家
【来源】https://www.ithome.com/0/758/402.htm
Views: 1