Google has released a new video generation model, Veo2, designed to compete with OpenAI’s Sora. Veo2 has significantly improved video realism and detailed performance, and has a deeper understanding of physical laws, human movement and subtle changes in expressions. This update not only improves video quality, but also enhances the model's ability to understand and execute user instructions, providing a more powerful tool for video creation. Through Google Labs' VideoFX platform, users can experience the powerful features of Veo2, generate videos with up to 4K resolution, and select video types, lenses, and movie effects. However, it should be noted that Veo2 is currently only available to users on the waiting list, and they need to apply through the Google form.
Google officially released its new generation video generation model Veo2, aiming to compete with OpenAI's Sora. Google says Veo2 exhibits greater realism and detail when generating videos. This update not only improves video quality but also enhances understanding of real-life physics and the nuances of human movement and expression.
The power of Veo2
Google says Veo2 has "a better understanding of real-world physics and the nuances of human movement and expression." It's available on Google Labs' VideoFX platform, but only on a waitlist basis. Users need to register via a Google form and wait for Google to temporarily grant access at a time of their choosing.
Google said that users can choose the type of video, lens and movie effects according to their needs, and Veo2 will be able to generate videos with up to 4K resolution.
In contrast, the original Veo is still available on Vertex AI, and videos generated by Veo2 will come with Google's metadata watermark SynthID to identify it as AI-generated content.
Despite Veo2's technical advancements, Google admits that the model occasionally exhibits "hallucinations" such as extra fingers, but overall the new version produces fewer hallucinations.
Google said that in its own internal testing, which measures "overall preference" (i.e. which videos viewers prefer) and "timely compliance" (how well a video matches instructions given by human creators), human raters preferred Veo , as opposed to the artificial intelligence models of Sora and other competitors.
Veo2 experience entrance: https://labs.google/fx/tools/video-fx
Introduction entrance: https://deepmind.google/technologies/veo/veo-2/
AI video generation still needs work
Although AI video generation technology is constantly improving, competition between Google and major players such as OpenAI remains fierce. Other smaller companies, such as RunwayML and Luma AI, are also rolling out new features to enhance their video generation capabilities.
Recently, RunwayML introduced advanced control features for the Gen-3Alpha Turbo model, while Pika Labs released Pika 2.0, allowing users to add their own characters to videos.
However, users’ trust in AI-generated videos is still insufficient. The release of Sora did not completely eliminate people's doubts, and some users reported that the results generated had physical and anatomical violations. Additionally, viewers at recent gaming awards shows expressed dissatisfaction with the “AI chaos” of AI-generated content.
Google said that many users have expressed strong interest in Veo2, especially some YouTube creators, who are already using VideoFX to create backgrounds for their short videos to save time.
Imagen3 updates
In addition to Veo2, Google has also updated its image generation model Imagen3, which further improves the realism and color vividness of the generated images. The new version of Imagen3 can more accurately represent a variety of artistic styles, from photographic realism to impressionism, to abstract and anime styles. At the same time, the model also performed better at following user instructions.
Highlights:
- Google releases Veo2 video generation model, claiming that its video quality is better than OpenAI’s Sora.
- Users can apply to use Veo2 through Google form, and can choose styles and effects when generating videos.
- The updated Imagen3 image generation model can better express various artistic styles and provide a better user experience.
All in all, the emergence of Veo2 marks another step forward in AI video generation technology, but it also reveals that the field still faces many challenges. Continuous advancement of technology and improvement of user experience will be the future development direction.