The year in Bard updates is starting today with Gemini Pro now available in “all supported languages and places.” Google is also adding an image generator to Bard and announcing “ImageFX.”
Gemini Pro launched in December and is now available in all of Bard’s existing 230+ countries/territories and 40+ languages. Google is also expanding the double-check feature beyond English to those supported languages. Accessed by tapping the “G” icon underneath a result, “Bard will evaluate whether there is content across the web to substantiate its response.”
Google Bard is also getting a free image generator powered by Imagen 2 that will be available “in English in most countries around the world.” Just enter a prompt/description, like “create an image of a dog riding a surfboard.” Google added a similar feature to the Search Generative Experience (SGE) last year.
Google recommends that you:
- Start your prompt with words like draw, generate, and create.
- Describe the image style you want. Here are a few examples: photorealistic, charcoal drawing, watercolor painting, cartoon illustration.
- Include a detailed visual description of what you want in the image. This can include the subject of the image, what the subject is doing, and the background or setting.
On the safety front, SynthID is used to “embed digitally identifiable watermarks into the pixels of generated images.” For example, Google Search’s About this image will tell you if an image was generated by Google AI. You also cannot generate images of named people, while:
Our technical guardrails and investments in the safety of training data seek to limit violent, offensive or sexually explicit content.
Google announced Imagen 2 in December with a focus on “high-quality, photorealistic outputs,” including accurate hand and face rendering while “keeping images free of distracting visual artifacts.” It powers text-to-image features in Duet AI for Workspace, SGE, Ads, and Google Cloud’s Vertex AI.
Imagen 2 has been trained on higher-quality, image-description pairings and generates more detailed images that are better aligned with the semantics of people’s language prompts.
It’s also available in Google’s new ImageFX experiment. This text-to-image tool is “designed for experimentation and creativity.” Starting with a text prompt, you can modify them with “expressive chips.” SynthID is in use, while generated images also include IPTC metadata.
Available through AI Test Kitchen, it joins TextFX and MusicFX, which has picked up model (MusicLM) improvements that result in higher-quality audio and fast generation. It’s available to English users in the US, Kenya, New Zealand, and Australia.
FTC: We use income earning auto affiliate links. More.
Comments