Individuals with an existing Google account have the opportunity to experiment with Google's newest AI-driven image generator instantaneously.
Google's most recent AI image creation tool, referred to as Imagen 3, is now accessible to the general public in the United States, requiring only a free Google account to experiment with. As pointed out by VentureBeat, Google covertly launched this model this week, publishing its research in a paper on Tuesday. This follows two months after the company initially announced the new model at Google I/O in May.
According to a post on Hugging Face, a machine learning platform, Google researchers shared, "We introduce Imagen 3, a latent diffusion model that generates high-quality images from text prompts. We detail our quality and responsibility evaluations. Imagen 3 surpasses other cutting-edge models in terms of performance at the time of evaluation. Additionally, we discuss issues related to safety and representation, as well as measures we employed to minimize potential harm caused by our models."
Based on some preliminary tests conducted by me, the model seems quite reliable. When requesting photorealistic images, the results are generally of high quality, boasting convincing attempts at realism that could deceive individuals on first glance. (I was particularly impressed with the image quality when I instructed the model to create a retro 35mm film look.) Imagen 3 also highlights significant aspects of the prompt that influenced the outcome, allowing you to adjust these if you're unhappy with the final image.
Credit: Jake Peterson/Google
Nonetheless, Imagen 3's outputs retains subtle hints of AI-generated images. In certain photos, hands have an odd number of digits, faces are distorted, and text makes little sense. (Although the model was able to accurately reproduce the "Coca-Cola" and "Canon" logos, infringing on trademarks.)
Google is not the only tech company releasing a new image model this week. X recently introduced a new beta version of Grok, its AI chatbot, and bringing with it, an image generator with seemingly minimal limitations. Users, specifically Michelle Ehrhardt from Lifehacker, have created a variety of content using Grok, from Taylor Swift wearing a MAGA hat to Pikachu wielding an AK-47.
Credit: Jake Peterson/Google
Conversely, Imagen 3 exhibits clear safeguards. When I attempt to use Google's image generator for controversial topics, it halts itself and politely directs me to Imagen 3's FAQ to comprehend why the prompt was inappropriate. It also refuses to generate copyrighted material, but can be tricked into yielding it with precise prompts. As mentioned earlier, I was able to reproduce logos, even managing to coax out trademarked characters like Mario and Pikachu, although they wouldn't participate in a shootout.
How to use Imagen 3
Credit: Jake Peterson/Google
Google's new AI image generation tool is free for any U.S. user with a Google Account to try out. To do so, navigate to ImageFX in Google's AI Test Kitchen. Simply sign in with your Google Account, and you're good to go, prompt away.
The initial release of Imagen 3, a groundbreaking AI image generation tool by Google, has sparked interest within the tech community. With Imagen 3, users can create high-quality images from text prompts, making it a notable addition to the tech field's arsenal of AI tools.
Furthermore, tech giants like Google and X are continuously pushing the boundaries of AI, as seen with Google's Imagen 3 and X's image generator in their respective new releases. These advancements are opening up new avenues for content creation, as demonstrated by the diverse images generated using Grok.
