close
close

How to try Google’s Imagen 3 AI Image Generator now

How to try Google’s Imagen 3 AI Image Generator now


Google’s latest AI model for image generation, Imagen 3, is now publicly available in the US. To try it out, all you need is a free Google account. As VentureBeat discovered, the company quietly unveiled its model this week and published its research findings in a paper on Tuesday. This comes two months after the company first announced the new model at Google I/O in May.

In a post on Hugging Face, a machine learning platform, Google researchers said: “We introduce Imagen 3, a latent diffusion model that generates high-quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss safety and representation issues and methods we used to minimize the potential harm of our models.”

After some initial amateur testing on my part, the model seems pretty solid. When requesting photorealistic images, the results are of a relatively high quality, with some good attempts at looking realistic that might fool people at first glance. (I was particularly impressed with the quality of the images when I asked the model to create a 35mm film look.) Imagen 3 also highlights certain parts of the prompt that affect the output, so you can adjust these if you don’t like the way the image turns out.

Request to create a 35mm look for a football match


Photo credit: Jake Peterson/Google

However, Imagen 3’s offerings still bear the telltale signs of AI-generated images. In some photos, hands have too many fingers, faces are distorted, and text doesn’t make sense. (Though the model was able to reproduce the “Coca-Cola” and “Canon” logos with trademark-violating accuracy.)

AI image of a Coke bottle


Photo credit: Jake Peterson/Google

Google isn’t the only tech company releasing a new image model this week. X recently released a new beta version for Grok, the company’s AI chatbot, an image generator that works with what appears to be very few limitations. Users (notably Michelle Ehrhardt of Lifehacker) have used Grok to generate everything from Taylor Swift wearing a MAGA hat to Pikachu with an AK-47.

Imagen 3, on the other hand, has obvious safeguards in place. When I try to prompt Google’s image generator with something controversial, it stops itself and politely refers me to Imagen 3’s FAQ to understand why my prompt wasn’t appropriate. It also refuses to generate copyrighted content, but can be made to produce it with the right prompts. As mentioned above, I was able to generate logos and even get it to spit out trademarked characters like Mario and Pikachu, although I couldn’t get them to engage in a gunfight.

AI-generated Pikachu


Photo credit: Jake Peterson/Google

How to try Imagen 3

Google’s new AI image generator is free to try out by any US user with a Google account. To do so, head to ImageFX in Google’s AI Test Kitchen, log in with your Google account, and then receive the appropriate prompts.

Leave a Reply

Your email address will not be published. Required fields are marked *