A new technology called TurboDiffusion claims it can speed up artificial intelligence A recently published paper shows up to 200x faster video generation at scale, without compromising overall image quality.
When the technology was tested on a consumer-grade system, it reduced the time required to generate a 5-second standard-definition video clip from more than 3 minutes to just 1.9 seconds, an acceleration of nearly 100 times. NvidiaRTX 5090 graphics processing unit, according to Beijing researchers Tsinghua Universitywith AI model developer Shengshu Technology University of California, Berkeleyin a paper published late last week.
In one experiment cited in the paper, the time it took to produce a high-resolution 5-second video on the same machine was reduced from 80 minutes to nearly 24 seconds, or about 200 times faster.
Currently, it takes users three to five minutes to generate a standard five-second video on Shengshu's Vidu platform. OpenAI The company's text-to-video model, Sora, also says it takes several minutes to render short clips.
TurboDiffusion highlights that Chinese researchers have made rapid progress in advancing AI video generation technology since OpenAI introduced Sora in February 2024.
According to industry analyst Kyung Hsu, the acceleration of video creation represents a major change, as the use of video generation models will no longer be a bottleneck in the production process.
