At the I/O 2024 developer conference, Google announced a remarkable AI technology breakthrough - the Veo model. The AI model's ability to generate a minute-long 1080p video based on text prompts marks a new milestone in video generation technology. Veo's release not only competes with leading models like Sora in the video generation space, but also showcases Google's innovative prowess in capturing visual style and editorial tweaks.
Demis Hassabis, head of DeepMind at Google, revealed in a virtual roundtable that the company is exploring Veo's potential for storyboarding and generating longer scenes, illustrating a new direction for future video production. The Veo model is built on Google's Imagen 2 series of image generation models, with significant improvements in resolution and video length compared to its predecessor.
While the source of Veo's training data isn't explicitly disclosed, DeepMind's Douglas Eck confirmed that some of the data may have come from YouTube and was in line with the creator's agreement. This approach has sparked discussions about data access rights and creators' rights. Eck said that Google will work with stakeholders in the film industry, music industry and other stakeholders to explore the future development of Veo, and gradually roll it out to a wider range of application scenarios.
The controllability of the Veo model is reflected in the understanding of camera motion and visual effects, as well as a certain mastery of physics, which enhances the realism of the video. Google has offered Veo trials to select creators, including well-known artist Donald Glover, and the market has responded positively to its positioning as a creative tool.
Not only is the AI model capable of generating a minute-long 1080p video based on text prompts, but it also showcases Google's innovative prowess in capturing visual style and editorial adjustments. The release of Veo marks a new milestone in video generation technology, and it has also sparked discussions about data usage rights and creators' rights.
The Veo model is built on Google's Imagen 2 series of image generation models, with significant improvements in resolution and video length compared to its predecessor. According to DeepMind's Douglas Eck, Veo's training data sources may include YouTube and are in line with the creator's agreement. Although this approach has achieved a technological breakthrough, it has also sparked discussions on the right to use data and the rights and interests of creators.
Google's use of YouTube data to train AI models has sparked discussions about data usage rights and creator rights. The New York Times reported in April that Google expanded its terms of service last year, in part because the company was able to use more data to train its AI models. Under the old terms of service, it was unclear whether Google could use YouTube data to build products other than the video platform. This is not the case under the new provisions, which significantly loosens the reins.
Google is far from the only tech giant that uses vast amounts of user data to train internal models. But what is sure to disappoint some creators is that Eck insists that Google has set the "gold standard" here, in terms of ethics. The solution to this challenge will be to get all the stakeholders together and figure out what's next, and Eck said we're not going to move quickly unless we take those steps with our stakeholders — we're talking about the film industry, the music industry, the artists themselves.
First of all, the technical foundation and training process of the Veo model deserve attention. According to DeepMind's Douglas Eck, Veo's training data may have come in part from YouTube, which has sparked discussions about data usage rights and creator rights. Last year, Google expanded its terms of service, allowing the company to leverage more data to train its AI models. This change has eased restrictions on the use of data to a certain extent, but it has also raised concerns about the protection of creators' rights.
Google isn't the only tech giant that uses user data to train internal models when it comes to ethical considerations. However, Eck insists that Google has set a "gold standard" when it comes to ethics. He proposed that the solution to the training data challenge is to bring all stakeholders together to discuss the way forward. This includes the film industry, the music industry, and the artists themselves, whose participation is critical to the future development of the Veo model.
The hands-on experience of the Veo model should not be overlooked. Veo's understanding of camera motion and visual effects, as well as his mastery of physics, enhance the realism of the video. Google has already offered Veo trials to select creators, including well-known artist Donald Glover, and the market has responded positively to its positioning as a creative tool.
However, the Veo model is not flawless. It illustrates the limitations of today's generative AI, such as the disappearance and reappearance of objects in video, as well as physical blunders, such as the impossibility of a car reversing. These issues indicate that the Veo model still needs further improvement and optimization.
In terms of market response, the gradual rollout of the Veo model shows that it can revolutionize the world of video production. But the development of this technology also comes with ethical challenges. How to balance innovation and creators' rights will be a problem that Google will need to face in the future. According to Eck, Google will work with various stakeholders to explore the future development of Veo and gradually roll it out to a wider range of use cases.
The release of the Veo model is not only a technological leap forward, but also raises important questions about data access, creator rights, and ethical challenges. As technology continues to advance and applications continue to expand, we expect Google to find the right balance between innovation and ethics to push video generation technology in a more mature and responsible direction.
With the gradual promotion of the Veo model, we should expect it to bring revolutionary changes to the world of video production. However, the development of this technology also comes with ethical challenges, and how to balance innovation and creator rights will be a problem that Google will need to face in the future.
In the future of video production, the Veo model illustrates a new era of personalized and automated content creation. With the advancement of technology, we can foresee a future in which the production of films and videos will no longer be limited to professional studios, but can be realized by anyone with creative ideas and text prompts. The democratization of this technology may lead to the democratization of content creation, but it also brings concerns about originality and copyright protection.
In terms of the call to action, Google and other tech companies need to work closely with creator regions, legal experts, and ethicists to develop clear guidelines and policies to ensure that the development of AI technology does not infringe on the rights and interests of individual creators. Issues such as copyright, ownership, and use rights for AI-generated content need to be clearer in terms of legal provisions and industry standards.
The advent of the Veo model is not only a big step forward in technology, but also a big challenge to the existing creative ecosystem. Google's role in promoting the development of video generation technology is also responsible for guiding this change towards a virtuous circle. We look forward to finding a balance between innovation and ethics to start a new chapter in video production together."