Google unveils Imagen 2 text-to-image generation for Vertex AI

Google Cloud Vertex AI logo.
(Image credit: Google Cloud / YouTube)

What you need to know

  • Google debuts Imagen 2 as the second upgrade to reach Vertex AI in Cloud as a piece of text-to-image generation software.
  • Imagen 2 can create "photorealistic" products for customers based on prompts with hopes of ridding users of the typical problems such bots tend to have.
  • Imagen 2 joins Google's Gemini Pro AI model, which Android app developers have recently gotten their hands on for integration.

Google unveiled its latest text-to-image generation software, Imagen 2, now "generally" available for Vertex AI customers. The company announced the upgrade in a Google Cloud post, stating the newest software brings six features with its debut.

The first involves "high-quality" imagery, as Google states Imagen 2 can create "photorealistic" output as it better understands what users intend to manifest.

This likely goes hand-in-hand with Imagen 2's text rendering support. The purpose of this is to give customers more control over their text-to-image prompts. Essentially, this support helps to ensure the root of a user's prompt remains consistent throughout the software's image generation, which is a problem Google adds is common in the industry.

Imagen 2 in Vertex AI will let businesses create multi-purpose logos, emblems, and lettermarks. The company adds users can take their creations and place them on products, business cards, clothing, and other unspecified surfaces.

Captions help the internet remain inclusive for everyone, which is why Imagen 2 allows users to create "descriptive, long-form captions" for whatever they create. Moreover, you can inquire about the image you've created with the AI chatbot, and it'll provide a tailor-made response.

Imagen 2 launches in English alongside support for Chinese, Hindi, Japanese, Korean, Portuguese, and Spanish in preview. Google states more languages are slated to be supported in early 2024. Users can also begin a prompt in one language but instruct the AI bot to output the response in another — likely useful for captions.

Google Gemini multimodal hero render

(Image credit: Google)

Google Cloud's inclusion of Imagen 2's influence doesn't come without safety parameters. The company adds that the software remains within the purview of its AI Principals with guardrails to ensure users cannot prompt the bot to create something harmful. Watermarks are also available through DeepMind's SynthID to protect creations made for businesses.

With Vertex AI customers getting their hands on Imagen 2, Google instructs those interested to reach out to their Google Cloud representative to join the Trusted Tester Program. Similarly, the image generation's documentation can provide additional information.

Imagen 2 now joins the Gemini Pro model in Vertex AI, designed to handle a "wide range" of tasks. Developers in Google's AI studio can also begin integrating the model into their Android apps. Doing so will enable their app to run off Google's data centers without requiring a completely new backend structure.

The company detailed Gemini's other two forms, Nano and Ultra, last week. The former is fit for on-device tasks and, as such, has rolled out on the Pixel 8 during its first feature drop. The latter, much larger for highly complex tasks, will get to shine early in 2024.

Nickolas Diaz
News Writer

Nickolas is always excited about tech and getting his hands on it. Writing for him can vary from delivering the latest tech story to scribbling in his journal. When Nickolas isn't hitting a story, he's often grinding away at a game or chilling with a book in his hand.