Google Launches Imagen 3 and Video Generative AI Models
Google has introduced two new generative AI models, Imagen 3 and Video (Veo), on its Vertex AI platform. These models can function independently or in combination, as reported by GSMArena.
Imagen 3 specializes in generating highly detailed images from text prompts, with improved handling of lighting and object details compared to previous versions. The model will be available to Vertex AI users starting next week. Google has showcased sample images and prompts created using Imagen 3 to demonstrate its capabilities. Additionally, corporate users can leverage the "Imagen 3 Customization" feature to integrate brand-specific styles or product characteristics into their outputs.
The Video (Veo) model can create animations from a still image and a text prompt or generate entirely new videos based on a textual description.
Google ensures the authenticity of AI-generated content by embedding DeepMind’s SynthID watermark in outputs from Imagen and Veo. This helps differentiate AI-generated content from real-world media. Furthermore, safety filters have been added to prevent the creation of harmful or inappropriate content.
Google stated that it has not used customer data to train these models and guarantees copyright protection for their outputs.
