New Image AI “Lexica Aperture” Generates Photorealistic Images


Photos and AI images are still easy to tell apart – but for how much longer? A new tool further blurs the lines between real and generated images.

Sharif Shameem, who created the vast AI image database Lexica Art, unveiled his new project: Lexica Aperture is an image generator that produces realistic images from simple text input in seconds.

Current AI image generators are not good at generating photorealistic images. Imitating various artistic styles in Midjourney, DALL-E and co. is already at a high level (much to the chagrin of some artists), but photorealism is still experimental. With the right prompts, however, you can get close.

Midjourney V4 demonstrated this very impressively, and DALL-E 2 already produces photorealism with the right prompt. No wonder AI image databases are preparing to compete with traditional photo directories, Shutterstock is starting cooperation with OpenAI, and some photographers are wondering whether their hobby or profession is over.

A d

Lexica Aperture offers a simple web interface

Lexica Aperture can be tried for free through the simple web interface after login. The settings are limited to adjusting the resolution between 512 x 768 pixels, the orientation scale between 4 and 13, and a switch to avoid double heads that often occur in AI Portraits.

Photo: Lexica Art Screenshot

Next to the prompt is a field for negative input of what should not be visible in the image. A few seconds after clicking “Generate”, Lexica Aperture spits out four results. Once completed, images can be resized to a maximum size of 3072 x 2048 pixels.

Stable broadcast formed with high resolution photos

Developer Shameem does not provide a deeper insight into the model’s architecture at this time. However, it is almost certainly a refined version of Stable Diffusion.

“Since the model is trained on high resolution photos, I found that a decent heuristic for Aperture is to think of it as a photorealistic 3D engine that you can request via text.” Shameem writes.

Lexica Aperture certainly won’t put photographers out of work just yet, even though the faces look more realistic than ever at first glance. Too often, closer examination reveals artifacts and inconsistencies that expose the image as the result of an AI and not a camera. Still, the model does a great job with stills, especially when simulating shots on older film.


Leave a Comment

Your email address will not be published. Required fields are marked *