Google AI has decided to bring back the human image generation tool that was previously disabled due to its production of historically inaccurate images.
Google revealed an upgrade for its Imagen 3 model in a blog post on Wednesday. The model was first released in May, and the latest version will soon start producing images of individuals for Gemini Advanced, Business, and Enterprise customers.
“The company announced that in the following days, they will begin to introduce the creation of images of individuals for their Gemini Advanced, Business, and Enterprise users, initially in English with a pre-release version.”
“Imagen 3 has allowed us to make substantial advancements in enhancing the user experience when creating images of individuals.”
Google reminded users that the AI image generator is not capable of producing photorealistic images of recognizable individuals, depictions of minors, or graphic and explicit scenes involving violence or sexuality, as of August 2024.
The statement clearly indicates that Google intends to prevent the creation of contentious images. The company elaborated that their restriction on “identifiable individuals”also applies to specific search queries that may result in images of prominent figures.
In response to complaints about factually and historically inaccurate images, Google disabled human image generation in Gemini back in February. Criticisms were directed towards Gemini’s depiction of certain individuals, including the US Founding Fathers and groups such as Nazi-era German soldiers, as people of color.
Some individuals also expressed dissatisfaction with the fact that Gemini does not generate images of people of various ethnicities. For instance, one user attempted to produce an image of a “blonde-haired, blue-eyed, Scandinavian woman,”but instead received an image of an “Indian woman with dyed hair”from Gemini.
Google issued an apology at the time for what it called “errors in certain historical image reproductions”made by Gemini. The company stated that the AI tool’s efforts to produce a diverse array of images fell short of expectations.
Shortly after Elon Musk acquired X, the image generation upgrade was implemented. This was in addition to a similar feature to Grok that was included in the acquisition. Furthermore, Midjourney made its image generator accessible to individuals without Discord accounts. While Stable Diffusion is still a viable option for local image generation, it requires a certain level of dedication.
Leave a Reply