Google is introducing its new basic text-to-image model, Image 3, to the Vertex AI platform, offering improved image generation capabilities, faster comprehension, and more realistic rendering of people. Initially revealed at Google I/O and available to select creators, Image 3 is designed to provide richer detail, reduced visual artifacts, and enhanced text representation within images. The model supports multiple languages, security features like SynthID digital watermark, and various aspect ratios.
Shutterstock is among the companies leveraging Image 3, with the platform seeing millions of images generated through the model since its implementation. Justin Hiza, VP of data services at Shutterstock, highlights the potential of Image 3 to expedite idea execution without compromising quality, with added security measures embedded in the content creation process.
Despite its advancements in Image 3, Google has not provided a timeline for when Gemini, its AI model generating images through multiple modalities, will resume functionality following concerns about inaccuracies. Google Cloud CEO Thomas Kurian emphasized the distinction between Image and Gemini, noting that they serve different purposes and technologies. Gemini is a multimodal model for reasoning across various information types, while Image focuses on high-fidelity text-to-image generation.
As Google continues to innovate in the AI space, questions remain about the future integration of Gemini’s imaging functionality and the company’s plans for the model. VentureBeat Transform 2024 is an upcoming event featuring industry leaders from OpenAI, Chevron, Nvidia, Kaiser Permanente, and Capital One, offering valuable insights into GenAI and networking opportunities over three days. Interested participants can register to attend and explore the latest advancements in AI applications across different industries.
Article Source
https://venturebeat.com/ai/googles-imagen-3-text-to-image-foundation-model-comes-to-vertex-ai/