Google LLC today announced that its Bard artificial intelligence chatbot is getting fully upgraded with Gemini Pro, the company’s most powerful large language model to date, which is also now available with image generation capabilities powered by its Imagen 2 model.

The company also introduced ImageFX, a new image generation tool, and an upgrade to MusicFX, which is an experimental text-to-music AI model.

Gemini Pro has already been available in Bard since December but only for a select subset of users in English. With this update, it will roll out globally for users in more than 40 languages and across more than 230 countries and territories.

Gemini represents Google’s most powerful LLM, capable of advanced text generation, answering questions, summarizing documents, conversational logic and coding capabilities. It comes in three different model sizes: the Pro, a smaller Nano designed for running on Pixel phones and mobile devices, and an extremely powerful Ultra, designed for enterprise services.

Alongside the upgrade, Google Bard will gain the capability to generate images using text prompts for users with the Imagen 2 text-to-image model. It’s the second generation of the Imagen model Google debuted in May 2022.

It can produce vivid, imaginative and photorealistic images from text descriptions written by users. The addition of image-generating capabilities to Bard brings it in line with Microsoft Corp.’s Bing Chat, which uses OpenAI’s DALL-E 3 to produce images from user chat.

“Just type in a description — like ‘create an image of a dog riding a surfboard’ — and Bard will generate custom, wide-ranging visuals to help bring your idea to life,” Jack Krawczyk, product lead for Bard, said in the announcement.

To promote the safe sharing of artwork produced by Bard, all graphics will be watermarked by SynthID, a tool developed by Google DeepMind researchers for AI-generated images that allows them to be identified. SynthID watermarks are invisible to the human eye but easily discerned by computer-aided tools.

Google’s new standalone ImageFX tool, also powered by Imagen 2, was added to the company’s AI Test Kitchen, a place where the company allows public access to experimental AI tools. Google also updated MusicFX, a text-to-music AI model that allows users to make songs.

ImageFX works just like any other generative AI artwork creation tool that allows users to input simple text prompts to produce images and then work with them by continuing to modify them with further prompts.

“People often discover new ideas through testing a range of prompts and concepts as they iterate,” Kristin Yim, product manager at Google Labs, said in the announcement. “To spur further creativity, ImageFX includes a prompt interface featuring ‘expressive chips’ that let you quickly experiment with adjacent dimensions of your creation and ideas.”

MusicFX uses Google’s MusicLM AI model that is capable of generating high-fidelity musical tracks from user text prompts or through humming a tune. Google introduced the text-to-music experiment last year and since then users have created more than 10 million tracks. It has now been upgraded to allow the creation of 70-second music loops and allows exploratory prompts through “expressive chips” to iterate on generated music.

“With feedback and improvements to our underlying MusicLM model, we’re enabling new capabilities like higher-quality audio and faster music generation,” said Yim.

Both ImageFX and MusicLM use SynthID to watermark their outputs so artwork and songs can be identified as AI-generated, Google said.

 Image: “Gemini Twins,” AI image generated by Google Bard