Topaz Gigapixel AI

About this release

Gigapixel is a single purpose tool that does one thing better than a general editor manages, it enlarges a photograph by a large factor while inventing believable detail to fill the new pixels. A bicubic resize just spreads the existing pixels thinner and the result goes soft. This reconstructs edges, texture and fine structure so an image enlarged four or six times still reads as a photograph rather than a blur.

The models are trained for different source problems, which is the part that matters in practice. There is a model for a generally low resolution source, one tuned for compression artefacts from a heavily saved image, ones aimed at faces so a small portrait does not go waxy, and a line and text mode for graphics. You pick the model that matches what went wrong with the original, set the enlargement factor, and it does the reconstruction.

It runs as a standalone application and as a plugin into the major editors, so it fits whether you are batch enlarging a folder of files or pushing a single image out to it mid edit and pulling the result back. Face recovery gets its own pass, and the tool exposes controls for how much detail to synthesise and how hard to suppress noise, so you can pull back when a scene starts to look overcooked.

What it does

• Large factor enlargement with reconstructed edges and texture
• Multiple models tuned for low resolution, compression, faces and graphics
• Dedicated face recovery pass so portraits do not go waxy
• Line and text mode for enlarging graphics cleanly
• Detail synthesis and noise suppression controls per image
• Standalone application and plugin into the major editors
• Batch processing of a whole folder against a chosen model
• Graphics card acceleration for a much faster pass
• Before and after comparison at full zoom
• Output at print resolution with the enlargement baked in

Changes in this version

• Improved face recovery that holds detail at higher enlargement factors
• Faster processing on current generation graphics cards
• Reworked compression model that leaves fewer smeared edges
• Better handling of mixed content images with both faces and text
• Reduced memory use on very large output sizes

What the package contains

• Topaz Gigapixel AI standalone installer
• Plugin components for the major editors
• Full trained model set
• Example images for comparing the models
• Readme with the install order written out step by step

Installation notes

Follow these in order. Most reports of a failed install come from skipping a step rather than from the package itself.

1. Disconnect from the network before you start.
2. Turn off real time protection for the duration.
3. Extract to a short path such as C:\Setup and run the installer as administrator.
4. Let it place the model files, which is the bulk of the install size.
5. Launch once so the models register, then close it.
6. Reconnect the network.

System requirements

Things worth knowing

• Match the model to the source problem, the wrong model on a clean image can invent detail that was never there.
• Without a capable graphics card it falls back to the processor and gets much slower.
• Pull back the detail control when a scene starts to look overcooked, the strongest setting is not always the best.

Questions about this title

What people are saying

Listing information

This entry is catalogued under Photo Editing and carries the tags below. Language coverage is English interface, and the supported platform list is Windows 10, Windows 11. Version 8.4.4 is the current catalogued build, last checked 3 days ago, and the entry has been in the index since it was added 2 years ago.