A new Google Gemini update brings custom Gems and better image generation

Google’s Gemini is rolling out exciting new features that were first previewed at Google I/O. These updates include the introduction of custom Gems for Gemini Advanced, Business, and Enterprise users, and the integration of the advanced Imagen 3 image generation model across all Gemini versions.

Custom Gems: personalized AI experts

Gems are an interesting feature that allow users to personalize Gemini by creating their own AI experts on any topic they desire. By providing instructions, a name, and engaging in conversations, users can craft Gems that serve as specialized assistants.

These Gems can be utilized for various purposes, such as brainstorming ideas, seeking advice on complex projects, or even generating creative social media captions. The ability to retain detailed instructions empowers Gems to streamline repetitive or challenging tasks, saving users valuable time.

If you are like me, you may have trouble fully grasping how Gems would be beneficial to use in your workflow. Thankfully, Google seems to understand that this new feature may need a bit of an intro to help users fully embrace it. To facilitate this, Google is launching a collection of pre-made Gems catering to different needs:

• Learning coach: Simplifies complex topics for easier comprehension
• Brainstormer: Sparks inspiration for various scenarios, from party themes to gift ideas
• Career guide: Assists in career development with personalized plans
• Writing editor: Provides constructive feedback on grammar and structure to enhance writing
• Coding partner: Enhances coding skills and supports project building

The rollout of Gems is underway for Gemini Advanced, Business, and Enterprise subscribers on both desktop and mobile devices, spanning over 150 countries and supporting most languages.

Imagen 3 sets a new standard for AI image generation

Additionally, Gemini’s creative image generation capabilities are getting a significant upgrade with the integration of the Imagen 3 model. This cutting-edge model empowers users to generate high-quality images using just a few words. Moreover, users can request images in various styles, such as photorealistic, oil painting, or claymation.

Imagen 3 boasts advanced features while upholding built-in safeguards and adhering to Google’s product design principles. It has demonstrated favorable performance compared to other image generation models across various benchmarks and the inclusion of SynthID (a tool for watermarking AI-generated images) further reinforces safety measures.

User control remains central to the creative process. If an initial image doesn’t meet expectations, users can simply provide feedback to Gemini, prompting the generation of a new image.

Expanded image generation capabilities

In the near future, Gemini will also introduce the generation of images featuring people, starting with an early access version for Advanced, Business, and Enterprise users in English. Google promises that technical improvements, enhanced evaluation sets, red-teaming exercises, and clear product principles have been implemented to ensure a positive user experience. When it comes to AI-generated images of people, however, there are a lot of issues that crop up quickly. This will most definitely be an interesting space to watch as these updates roll out later this year.

There’s no doubt that this latest update is a pretty significant step forward for Gemini. The introduction of custom Gems and the integration of Imagen 3 offer users some serious personalization and creative possibilities. It certainly feels like the AI race is continually shifting and moving under our feet, and with each additional service, Google is focusing on use cases that make sense and are actually helpful to the user. We’ll be on the lookout for these new features to arrive soon.