Google has recently implemented the Image 3 rendering engine on Gemini, opening the doors of rendering to any user of this AI. This AI can be used online, but its potential on mobile seems pretty crazy to me. In my case, I have Gemini instead of Google Assistant, so I just press the power button to activate it.

By turning on Gemini directly, I can give it instructions with my voice and I don’t need to type commands. Pressing a button and telling it to “create the image of a Mercedes SLR McLaren on the driveway” and creating that image perfectly was something I wasn’t expecting.

Image 3 is the new version of the artificial intelligence model that Gemini will use to create images from now on. This is the latest version and according to Google it has improved these three key features over previous versions:

• Images have much more detail and fewer artifacts.
• Images are much more realistic than before.
• It is now possible to create images with much more style.

Using Image 3 on Gemini is very simple. Just download the Gemini app from the Play Store and create the prompt. It can be written down, but it is much more comfortable to do it out loud.

Being a lover of good cars, I asked Gemini to create the image of a Mercedes SLR McLaren on the speedway. I chose this model because it is quite special in terms of design and to my surprise, he recreated it -almost- perfectly. We can still say that this is an AI-generated image, but for a photo created in five seconds with my voice, it’s not that bad.

In the second image, I created something simpler; A dog and a cat walking together. The bokeh it creates is simply gorgeous and the image is brutal in terms of sharpness.

To complicate things a little more and see how you interpret the styles, I asked you for a backlit CyberPunk-style rendering of a street in New York. The result is again surprising.

At the end of the holiday I asked you for a picture of an iMac with the word ‘Xataka’ on the desktop and wallpaper. Generating text is a problem for many free AIs. This is not the case in Figure 3. There’s no problem creating brand logos either, although it’s not that good here.

Image | Xataka

in Xataka | 19 pages and services for creating images from scratch using AI