Google Updates Gemini’s AI Image-Creation Model

(Reuters) – Alphabet’s Google (NASDAQ:GOOGL) announced on Wednesday that it has updated its Gemini AI image-creation model and will soon enable the generation of visuals of people after a months-long pause.

Background

In February, the company halted its AI tool for creating images of people due to inaccuracies in some historical depictions produced by the model. This led to user criticism as the AI sometimes returned erroneous historical images.

Improvements Made

Google has taken steps to enhance the product by adhering to its “product principles” and conducting simulations to identify weaknesses. The new feature will initially be available to paid users of the Gemini AI chatbot, starting in English. The rollout will expand to include additional users and languages over time.

Model Enhancements

The company confirmed that it has improved the Imagen 3 model for better representations of people. However, it will avoid generating images of identifiable individuals, children, or explicit content.

Competition

Other AI chatbots that offer image generation capabilities include OpenAI’s Dall-E, Microsoft’s (NASDAQ:MSFT) CoPilot, and recently xAI’s Grok.

New Features for Subscribers

In addition, Google stated that in the coming days, subscribers to Gemini Advanced, Business, and Enterprise will have access to customized chatbots called “Gems,” which can be tailored for specific tasks. Users will be able to provide detailed instructions to create a Gem, streamlining repetitive use cases.