On Tuesday, Google unveiled two new artificial intelligence (AI) features for its Gemini platform. The tech company, headquartered in Mountain View, announced the introduction of Canvas, an interactive tool designed for users to collaborate with AI on various document and coding tasks. Additionally, the Audio Overview feature, previously available only on NotebookLM, will now be accessible through Gemini. This function allows users to produce engaging, podcast-style audio discussions based on documents, slides, and detailed research reports. These features are being gradually rolled out worldwide for Gemini Advanced subscribers as well as users of the free tier.
Google Introduces New AI Capabilities to Gemini
In a recent blog post, Google outlined the incorporation of these two innovative features into Gemini. This announcement comes after the previous launch of the Deep Research tool, which generates comprehensive reports on complex subjects, along with the iOS-exclusive lockscreen widgets. The newly announced features, Canvas and Audio Overview, will be available across both the web and mobile applications of Gemini.
Canvas offers an interactive environment on Gemini that enables users to collaborate effectively with AI on specific projects. A new Canvas button has been added next to the Deep Research option in the application’s interface. By selecting this feature and uploading relevant documents or lines of code, users can create a sandbox where the AI generates a first draft based on their input. Subsequently, the user can edit this draft and refine the content with assistance from the chatbot.
Currently, Canvas is tailored for documents and coding tasks. When working with documents, users need to upload a file and provide a prompt while the Canvas option is active. For instance, they might request, “Create a speech based on these classroom notes,” prompting AI to generate a draft in the sandbox-like interface. Users can then manually edit or highlight sections, requesting the AI to modify the tone or regenerate specific content based on their feedback.
In terms of coding, users can ask the AI for code based on their prompts. With Canvas, they can view and preview scripts and web app prototypes, allowing for a visual representation of the code, currently limited to HTML and React. After this preview, users have the option to request adjustments to input fields or calls-to-action, along with seeing the updated preview. This functionality is reminiscent of OpenAI’s Canvas feature, which is exclusively available on the web.
