Mediatek’s real-time AI image generation is pure magic
If you’ve been online in the past year or so (we know you have), you’ve probably seen some jaw-dropping examples of artificial intelligence generating images based on text prompts. Typically, it takes some time for the image to be generated, but what if this could happen instantly, as you type?
Turns out, it can. At this year’s Mobile World Congress (MWC) in Barcelona, chip maker Mediatek showcased several impressive AI demos. I’ve seen a couple, but the one that impressed me the most was the demonstration of a SDXL Turbo, an AI engine that can dynamically create images in real time based on text prompt.
Mediatek wasn’t kidding about the real time bit. I’ve tried it out, and the image would instantly change based on what I was typing. I started with “red,” which resulted in an image of a red tapestry. As I added the word “cat,” Mediatek’s demo device immediately switched the tapestry with an image of a reddish cat. Whatever I typed next, I’d receive a quite usable rendering of it, practically immediately.
We’ve seen something similar on Samsung’s Galaxy 24 Ultra and its AI-based wallpaper generation, though the Mediatek demo I’ve seen creates images even faster, literally as you type.
In another demo, a text prompt would result in an AI-generated video, though that one wasn’t instant; it took roughly 30 seconds for the video to appear. One can easily imagine, however, that it’s just a matter of time before we can “direct” our own video by typing in what we want to see, with the AI generating results fast enough to appear instantaneous.
The demo units I’ve seen are using the latest generation of the APU (AI Processing Unit) hardware that’s present in Mediatek’s Dimensity 9300 smartphone chip, and Mediatek claims the demos were running entirely on-device. Like it or not, this means that the type of technology I’ve just seen might come to widely available smartphones soon.
For more from Mobile World Congress, Mashable’s got you covered.