Google’s AI services are receiving boosts across the board, with AI image generation arriving in Google Bard, powered by a new Imagen 2 technology that Google promises is more lifelike (and weirdness-free) than ever. And just days after the internet voted its Gemini Pro chatbot technology as nearly the world’s best, it’s arriving pretty much everywhere Bard is offered.
Google is also releasing a dedicated AI art tool, known as ImageFX, to go along with its existing MusicFX and TextFX tools, inside its AI Test Kitchen beta program.
Overall, it’s a multifaceted improvement for Bard, both from a quality standpoint and the number of services Google is integrating AI into. Although Google launched its Gemini Pro upgrade last December, the chatbot was available to just a subset of users. (Although Bard is the AI service, Gemini and Gemini Pro are the underlying conversational models Google developed.) Now, Google says that Gemini Pro is available in all supported languages, countries, and territories.
Google hasn’t disclosed many of the parameters used to create Gemini Pro, though it’s just one of three related models. A smaller Gemini Nano (with up to 3.5 billion parameters) is in use on Google’s latest Pixel phones. A more powerful version, known as Ultra, is “coming soon,” according to Google chief executive Sundar Pichai’s Tuesday comments to investors — along with an undisclosed subscription model attached to it. A Google representative declined to comment further.
Tellingly, however, a HuggingFace “leaderboard” of chatbots, rated by the public, puts Bard’s new Gemini Pro model just behind the “Turbo” iteration of ChatGPT 4 as the second-best AI chatbot — meaning you should try it. Bard is free, with unlimited conversations.
What Bard hasn’t done previously, however, is offer in-line generative AI art. Microsoft’s Copilot does. Google Search’s SGE technology does as well, but only after signing up for the AI Test Kitchen program that enabled MusicLM and similar experiments. (Google also tweaked its AI music-generation derivative, MusicFX, making samples both higher quality and faster to generate.) Now, you’ll be able to ask Bard for an image and it will draw it, for free.
ImageFX is Google’s latest generative art tool that works separately, but also taps Imagen 2’s artistic capabilities. Google has always lagged a bit behind Bing Image Creator (now a function of Microsoft Designer) as well as more general AI art tools like Midjourney in standalone AI art apps, but Google has now caught up.
Now, it appears that ImageFX will allow you to create your own standalone AI art, along with the ability to quickly modify it using what Google calls “expressive chips” that will allow you to quickly change the style. All images generated using ImageFX include metadata that identify the images as AI generated, Google said, “giving people more information whenever they encounter our AI-generated images.” ImageFX images won’t just be digitally watermarked using a Google tool known as SynthID; they’ll also show up in Google search as AI-generated.
The Imagen 2 technology “delivers our highest-quality images yet, as well as improvements in areas that text-to-image systems often struggle with, such as rendering realistic hands and human faces and keeping images free of distracting visual artifacts,” according to a statement by Eli Collins, vice president of Google’s DeepMind.
Google underwent “extensive adversarial testing and red teaming to identify and mitigate potential harmful and problematic content,” Collins added, perhaps obliquely referencing the recent sexualized, AI-generated images of Taylor Swift that were reportedly generated via Bing Image Creator and then published on Twitter.