Google is reopening the ability to generate images with its Imagen generator (Photo: Google).
The company is releasing a new version of its Imagen image generator that will allow users to re-create images of people. Google has had a lot of problems with this tool in the past.
Starting this week, the Imagen 3 model will be available to all Gemini users, but the ability to create images with people will first be available to subscription holders. English-speaking users of Gemini Advanced, Business and Enterprise will be able to use this feature in the coming days.
Google says the new version of the AI model sets new standards for image quality and doesn’t have the problems associated with creating historical images that the company faced earlier this year. The new restrictions will also prevent the creation of photo-realistic images of famous people who are easily recognizable.
The tool now allows you to adjust the created images with additional hints. They will help correct the result if it does not meet expectations. In addition, Imagen 3 includes Google’s SynthID tool for adding watermarks to images to make it clear that they are created by artificial intelligence.
Gemini Experiences Senior Director of Product Management Dave Citron notes in a blog post that images created with Imagen 3 may not be perfect. At the same time, he assures that the company will continue to listen to user feedback and improve the product.
What is known about the Imagen 3 scandal from Google? In February, Google launched an image generator in its Gemini AI chatbot. Users quickly discovered that the tool produced historically inaccurate images, such as black-skinned US “founding fathers” and Vikings, and had significant problems generating images of light-skinned people. When this information became known, Google announced that it was restricting access to the newly implemented Gemini feature.
In February, it was announced that Google’s Gemini image generator would be down for several weeks.
Google co-founder Sergey Brin publicly admitted that the company did not do enough to test the quality of its image generator before launching it in the Gemini chatbot.
