Gemini apps are going to get two new advanced capabilities, Google announced on Wednesday. The Mountain View-based tech giant’s in-house artificial intelligence (AI) chatbot will receive the AI agent Gems and image generation capabilities of the recently released Imagen 3 AI model. While the former will only be available to the paid users of Gemini, the latter will be shipped to all users, including those on the free tier. However, those using the free version might see some added limits to image generation.
Google made the announcement to integrate Gems and Imagen 3 into the Gemini apps in a blog post. Both features were first previewed at the Google I/O earlier this year. Notably, Gems has already been rolled out and will be available for Gemini Advanced, Business and Enterprise users. The company said the Imagen 3 features will be shipped in the coming days to Gemini, Gemini Advanced, Business, and Enterprise users.
Gems are essentially miniature versions of the chatbot with a limited dataset. They can be customised to focus on a specific set of topics, which enables the AI model to generate more specific and accurate information. Google said, “With Gems, you can create a team of experts to help you think through a challenging project, brainstorm ideas for an upcoming event, or write the perfect caption for a social media post.”
Users can also add specific instructions to a Gem to refine the responses further. Once the feature is available to users, they will also find a set of pre-made Gems created by Google. These include Learning coach, Brainstormer, Career guide, Writing editor, and Coding partner. Gems will be available in multiple languages on desktop and mobile devices in more than 150 countries.
Imagen 3, the company’s latest image generation AI tool, is also being rolled out to Gemini apps. It can generate images in different styles, such as Nikon DSLR, GoPro style, wide-angle lens, and more. Google says it can also generate “photorealistic landscapes, textured oil paintings, or whimsical claymation scenes.”
One significant upgrade with Imagen 3 is that the AI model will also let users generate images of people, something which was removed after many users noticed Gemini was generating biased and harmful images involving people. To reduce the risk of deepfakes, the company says it has added inbuilt safeguards. Further, SynthID has been used to watermark the images as generated by AI.
While the company did not specify, it hinted that Imagen 3 capabilities may also include inline editing of the generated images. However, it appears the editing can only be done using text prompts. Notably, Google says Imagen 3 will not “support the generation of photorealistic, identifiable individuals, depictions of minors or excessively gory, violent or sexual scenes.”