KYIV, UKRAINE, [Date] – Those of us who can describe images faster than we can draw them have a fascinating new tool at our disposal. Image-enhancement pioneer Image Upscaler is proud to announce the debut of its new text-to-image service, available free at https://imageupscaler.com/ai-image-generator/.
The AI Image Generator allows anyone to produce a unique digital image simply by entering a few keywords and choosing an artistic style. A convolutional neural network handles the rest, producing an entirely AI-generated PNG in less than a minute. From photorealistic mountain vistas to cats a la Van Gogh, the service can produce compelling images based on nearly any request thrown at it.
“We built the AI Image Generator to be a freeform source of inspiration for anyone interested in graphic arts,” says Image Upscaler [Title] Evgen Gunya. “Creatives sometimes need to look at things from a slightly different angle, and that’s exactly what our image generator does. For the rest of us, it’s just fun. We had a great time testing our system before we introduced the service to the public.”
He might be understating the sheer amount of work that went into the image generator. To produce compelling graphics based on freeform text, Image Upscaler made innovative use of diffusion models: neural networks capable of creating shades of meaning or visual representation between items on which they have been formally trained.
Generative neural networks have been around for some time, but diffusion models have only been widely deployed for a few years. A faster and more efficient take on the idea, latent diffusion modeling, was described earlier this year by researchers at the Ludwig Maximilian University of Munich.
The AI Image Generator uses latent diffusion modeling to interpret text inputs, relate them to the huge repository of images on which it has been trained, and create new images in response to each new request. At times, the results are brilliantly true to life. At other times, the network seems to take a bit of artistic license.
“Neural networks are still nowhere near as complex as the human brain, and our network still doesn’t capture the nuances of human expression with photorealistic accuracy,” notes Gunya. “Some of the results are striking, some take us into the uncanny valley, and others are so highly stylized that you can easily imagine a human artist adding little distortions and adjustments to real-life subjects.”
The AI Image Generator joins Image Upscaler’s growing suite of professional-grade image-manipulation tools, including background removal, enhancement, deblurring, and the company’s flagship upscaling service. All are powered by AI, and all benefit from Image Upscaler’s continued efforts to extend the usefulness of today’s most advanced technologies.
To learn more about the AI Image Generator, or about Image Upscaler’s other services, please contact Kateryna Yakymenko at kayte.yakymenko@imageupscaler.com.
