Google Cloud customers now have private preview access to Veo, Google's AI video generation model, through Vertex AI. This allows companies like Quora and Mondelez International to leverage Veo for content creation.
Veo's Capabilities and Improvements
Veo generates up to six-second 1080p video clips from images and prompts. It supports various styles, edits, and higher resolutions (720p) in landscape and portrait orientations. While proficient in VFX and physics, Veo, like other AI models, has limitations with object consistency and accuracy. Similar to advancements in other AI models, Google continues to refine Veo.
Training Data and Risks
Veo's training data includes various video-description datasets and potentially some YouTube content. Google ensures data curation for safety and security. Concerns about data sourcing and copyright infringement, similar to those raised with data brokers, exist. Google offers prompt-level filters and indemnity for eligible users to mitigate these risks.
Veo's Integration and Future
Veo has been integrated into Google Labs and YouTube Shorts, allowing creators to generate backgrounds and short video clips. Google employs SynthID watermarking to address deepfake concerns. While Google explores partnerships, the potential impact of AI on creative jobs, as discussed in relation to industry lawsuits, remains a concern. Veo's general availability and expansion to other platforms are yet to be determined.