Friday, May 20th, 2022 Posted by Jim Thacker

Check out Gigapixel AI 6.1

Originally posted on 9 May 2022. Scroll down for news of the Gigapixel AI 6.1 update.

Topaz Labs has released Gigapixel AI 6.0, the latest version of its AI-trained software for increasing the resolution of photos or rendered images without softening or introducing noise into the result.

The update adds native support for Apple’s M1 processors, reducing processing times on M1 Macs by around 40%; and an optimised 16-bit version of its Standard AI model to improve performance on Windows.

Full disclosure: the update was actually released last month, but we didn’t manage to cover it at the time.

Machine-learning-trained tool enlarges images without sacrificing detail
Gigapixel AI is designed to enlarge digital images while preserving their sharpness, using machine learning techniques to fill in the missing details.

Its algorithms are trained using a data set of “millions of photos of all subjects and styles”, and can enlarge photos by “up to 600%” before image quality starts to drop visibly.

The software supports batch processing, making it possible to run it on entire image libraries.

Works for renders as well as photos
Although it’s marketed primarily at photographers, Gigapixel AI can also be used on rendered imagery – and, as of Gigapixel 5.3, has a dedicated AI mode for CG images and scanned artwork.

As well as simply upscaling low-res renders to generate high-res images more quickly, it can be used to remove residual noise from renders by upscaling them in Gigapixel, then reducing them to their original size.

While there are a number of open-source libraries and free online services for upresing images, its supporters claim that Gigapixel achieves better-quality results.

New in Gigapixel AI 6.0: native support for Apple M1 processors
The main change in Gigapixel AI 6.0 is native support for Apple’s M1 processors.

According to Topaz Labs’ blog post, native M1 support cuts software startup times by 47% over Rosetta emulation, previously used on new Macs; and cuts image batch processing times by 44%.

The update also introduces optimised 16-bit backends for the software’s Standard AI model, which should improve performance on Intel Macs, and on Windows machines processing images on the GPU.

Workflow and UX improvements include increased tonal and colour contrast in the UI; and users now have the option to download all of its AI models on installation, to remove the need for manual downloads later.

Updated 20 May 2022: Topaz Labs has released Gigapixel AI 6.1.

The update introduces a new Face Recovery model for upscaling portrait images. It preserves detail on skin and eyes better than the previous Face Refinement model, avoiding Uncanny Valley-like effects.

In addition, Topaz Labs has changed the machine learning library used on Nvidia GPUs from DirectML to TensorRT, resulting in “significant performance improvements” on GeForce 10, 20 and 30 Series GPUs.

In the firm’s own tests on a current-gen GeForce RTX 3090 GPU, a 400% upscale of a 4,092 x 2,732px JPEG image was over twice as fast using TensorRT than using DirectML.

Pricing and system requirements
Gigapixel AI 6.1 is available as a standalone application for Windows 10+ and macOS 10.14+; and as a plugin for Photoshop and Lightroom Classic, on both Windows and macOS. New licences cost $99.99.

The software is OpenGL-based, so it should work on any manufacturer’s hardware, but requires a reasonably powerful GPU.

Read a full list of new features in Gigapixel AI 6.0 in the online release notes

Visit the Gigapixel AI product website