Stable Diffusion Rolls Out ‘Upscaler’ That Ups Images’ Resolution By 4X

Wish you could have a higher-quality version of a profile picture you set years ago? Thanks to artificial intelligence, you can get it at quadruple the resolution.

Stability AI, most known for building the popular text-to-image generator Stable Diffusion, has somewhat returned to the fundamentals with an open-source ‘4x Upscaler’ that increases the size of any image by up to 400% “while maintaining—or even improving—their level of detail.” While image upscalers aren’t revolutionary, their usefulness will be boosted now that AI-generated images—many of which are bogged down by resolution limits—are thrown into the mix.

The upscaling tool is being rolled out as an API and can apply to any photos or images, not just those generated on Stable Diffusion.

The 4x Upscaler amplifies image resolution in 20 to 40 seconds. In the same release, however, Stability AI says it’s also prepared a “super-fast” open-source model, called the ‘Real-ESRGAN’, that can double the resolution of 512 x 512-pixel images in half a second.

The 4X version takes a bit more time because it has a heightened perceived level of detail and yields clearer, richer results, notes the organization.

Stability AI says the two upscaling tools have also been added to the latest versions of Stability for Photoshop or Stability for Blender. Time to give your low-res photos or artwork a glowup!

[via DIYPhotography, PetaPixel, Stability AI, images via Stability AI]