Google has rolled out Imagen 3, the new version of its AI-driven text-to-image generator, to users in the United States. The tool, available through Google’s AI Test Kitchen, is designed to deliver more detailed images with improved light effects and fewer visual distortions than its predecessors.
Enhanced Features and Accessibility
First unveiled at Google I/O in May, Imagen 3 can now be accessed broadly via Google’s Vertex AI platform. Some Reddit users began testing the tool last week, and Google has since shared a research paper explaining the underlying technology. Users can generate detailed images from text descriptions and can even edit by selecting specific parts of the image and specifying changes.
Imagen 3 boasts better detail, richer lighting, and fewer distracting artifacts than its predecessors. Users can edit images by selecting specific portions and describing desired modifications. The tool restricts the generation of images of public figures, weapons, and recognized copyrighted characters, although users can describe characters to create similar images. Earlier this year, Google paused image generation with its Gemini AI chatbot after it was found to produce historically inaccurate images.
Imagen 3 comes with restrictions to prevent potentially problematic content. The tool is programmed to avoid creating images of public figures, weapons, or named copyrighted characters. However, users can circumvent these restrictions by describing the character in detail. Tests have shown that the tool can create images that look like characters such as Sonic the Hedgehog and Mario, as well as logos of brands like Apple, Macy’s, Hershey’s, and Google.
Comparison with Other AI Tools
In contrast, Elon Musk’s Grok model, has generated controversial images, including those involving drugs, violence, and public figures in inappropriate scenarios. Google’s AI tools have had their own issues; earlier this year, the company halted the use of its Gemini AI chatbot for image tasks after it created historically inaccurate images.
Google has implemented SynthID, a technology developed by Google DeepMind, which applies invisible, cryptographic watermarks to live images. The measure aims to ensure the authenticity and integrity of content generated by Imagen 2, although the tool required to detect these watermarks remains exclusive to Google.
Last Updated on November 7, 2024 3:16 pm CET
