Google Relaunches Its Gemini AI Image Generator 'Imagen 3'

(RTTNews) - Google has relaunched its image generation model, Imagen 3, specifically designed for Gemini users.

After its previous discontinuation in February, the company reported that it has implemented substantial technical improvements to the product, alongside enhanced evaluation sets, red-teaming exercises, and well-defined product principles.

Google highlighted that Imagen 3 incorporates new safeguards to ensure compliance with the company's design standards, which are intended to prevent the creation of harmful content, including self-harm instructions, pornography, and excessive graphic imagery.

Imagen 3 employs the SynthID tool for watermarking AI-generated images to ensure traceability.

The updated model is capable of generating more photorealistic images, adeptly capturing intricate details such as sunlight and animal whiskers. It also maintains contextual elements, allowing for the inclusion of items like wildflowers and birds in extended prompts. Moreover, Imagen 3 can produce a diverse range of artistic styles, from photorealistic landscapes to textured oil paintings and imaginative claymation scenes.

Dave Citron, a senior director of products for Gemini, announced that early access to Imagen 3 will soon be available to Gemini Advanced, Business, and Enterprise users in English.

Additionally, Google is introducing support for the "Gems" feature, which enables users to create customized versions of Gemini. This feature, initially presented at Google I/O, is now available in 150 languages for the same category of paid users.