Google Introduces Veo and Imagen 3 to AI Platform


Two new Google generative AI models, Veo and Imagen 3, are now available on the cloud hosting service Vertex AI. Veo generates videos, and Imagen 3 creates photos.

Google proposes these models could be used in marketing, e-commerce, and more.

Veo builds video from text or image prompts

After companies like OpenAI experimented with video generation to mixed success earlier this year, infrastructure and processing power have caught up and made video generation more practical. Just this week, Amazon introduced a video-generation model called Amazon Nova Reels on AWS.

Veo creates videos from text or image prompts and can add elements to existing videos. Google addressed common challenges in video-generation AI, such as maintaining continuity and avoiding unnatural-looking movement. This effort followed OpenAI’s February Sora demonstration, which highlighted issues such as bizarre movements, including surreal examples featuring wolves.

Veo can generate images over a minute long at 1080p resolution, and longer videos can include multiple scenes generated from sequential prompts.


Veo can replicate different art styles, as shown by this video, generated entirely by the model. Image: Google DeepMind

Veo is available now in private preview. To get started, account holders can contact their organization’s Google Cloud account representative.

Imagen 3 creates photorealistic or animated-looking images

Imagen 3, released in Vertex AI in a private preview in June, will be generally available “starting next week.” Imagen 3 can create realistic or stylized images from text prompts. Google said the detail, lighting, and artifact reduction have improved over Imagen’s previous generation.

For marketing, Google highlighted that organizations can incorporate their own “brand, style, logo, subject or product features” into AI-generated images. That customer data is not fed back into the model to train newer iterations of Imagen.

Google Introduces Veo and Imagen 3 to AI Platform
A company can insert its own product images into Imagen 3’s AI-generated content. Image: Google Cloud

Google has offered a guide to working with Imagen 3 for developers. For security, Imagen 3 comes with digital watermarking, content filters, and data governance safeguards.

SEE: These video production templates for Google Sheets and more could make your workflow run more smoothly.

Companies offer more generative AI options, explore ROI sources

Google Cloud has sold Veo and Imagen 3 on Vertex AI to various high-profile clients, who have praised the tools for rapid iteration of content using generative AI. However, generative AI giants and their customers sometimes struggle with how to derive value from the tools.

Meanwhile, Google claimed that 74% of organizations with generative AI investments see ROI. However, the reception to AI-generated content hasn’t been universally positive: Coca-Cola’s November AI-generated video ad created controversy, with some consumers reacting negatively to what they perceived as a lack of creativity or authenticity.



Source link