Google Launches AI Image Generation Model Imagen 3 and Video Model Veo

Google finally launched its latest artificial intelligence (AI) image and video generation models on Wednesday. Both of these AI models were unveiled at Google I/O as the tech giant's latest advancement in generative AI. Now, more than six months later, the Mountain View-based company has introduced it on Vertex AI for its enterprise clients. Notably, while Imagen 3 was not available as a standalone platform so far, it was being used in several platforms and tools such as Google Docs, Gemini, and an experimental tool called GenChess.

Google Imagen 3, Veo AI Models

In a blog post, the tech giant announced the introduction of the two new AI models in Vertex AI. Google's Vertex AI platform is a managed machine learning (ML) platform on Google Cloud that allows developers and enterprises to build, deploy, and manage AI models. It is similar to Amazon Bedrock and Microsoft Azure and offers integrated tools and solutions for AI workflows.

The tech giant stated that the Veo video generation model is now available on Vertex AI in private preview and businesses can generate videos using text or image prompts. On the other hand, Imagen 3 will be made available starting next week. It takes text prompts and enterprises can use it to generate images that reflect their brand style and logos.

Coming to the capabilities of Veo, Google says it can generate high-quality videos based on either text or image prompts. The generated videos can be in a wide range of cinematic and visual styles. Developed by DeepMind, the AI model is said to have high prompt adherence and can generate consistent footage of objects and people and even capture movements realistically.

Imagen 3, which will be available in Vertex AI starting next week, can generate photorealistic images in a wide range of styles. Calling it “our most capable image generation model yet”, Google stated that the image generation model can understand natural language prompts and users do not have to describe the technical elements to get the desired result.

The Imagen 3 AI model will also be available with editing tools for inpainting and outpainting. Companies can also infuse their brand's colours, styles, logos, and other elements in the generated images.

For privacy and safety, the tech giant has added several tools. SynthID, the watermarking technology developed by DeepMind, will be embedded into every image and frame of video that these AI models produce to combat instances of deepfakes and misinformation. Google also stated that the AI models will not be trained on customer data and the tools will operate following Google Cloud's data governance and privacy controls.