Topaz Gigapixel AI Pro

About Topaz Gigapixel AI Pro

Enlarging a photograph has always been a losing proposition. Traditional resampling has no information to work with, so it interpolates between existing pixels and the result is soft, mushy and obviously stretched. Gigapixel takes a different approach: models trained on enormous quantities of image pairs reconstruct plausible detail rather than averaging what is already present.

In practice that means an old scan, a heavily cropped frame or a small file from a phone can be taken up several times its original dimensions and still hold together at print size. Edges stay defined, texture in fabric and foliage survives, and the flat plastic look that gives away an upscale is largely absent when the right model is chosen for the material.

Several models ship because different source material breaks in different ways. There are models for standard photography, for compressed images carrying block artefacts, for lines and text, for low resolution originals, and a dedicated face recovery pass that reconstructs facial detail on subjects too small in frame for the main model to handle convincingly.

The Pro build adds the working side. Batch processing runs a whole folder against a chosen model unattended, the command line interface allows the whole thing to be scripted into an existing pipeline, and the plugin bridge sends a frame from a host editor and returns it in place. Output preserves colour profile and metadata, with control over sharpening and noise handling before export.

What is included in this build

• Model based enlargement reconstructing detail rather than interpolating
• Separate models for photography, compressed sources, lines and low resolution originals
• Dedicated face recovery pass for subjects too small in frame
• Batch processing across an entire folder unattended
• Command line interface for scripting into an existing pipeline
• Plugin bridge that returns a processed frame to a host editor in place
• Side by side comparison view with synchronised zoom before committing
• Colour profile and metadata preserved through the export

Changed in version 8.4.4

• New model producing better results on heavily compressed source images
• Faster processing on machines with a current generation graphics card
• Face recovery improved on profile and partially obscured subjects
• Batch queue now resumes correctly after an interruption
• Reduced memory use on very large output dimensions

Installation walkthrough

These steps assume a clean machine. If a previous release of the same application is present, deal with that first, since an overlapping install is the most common cause of a setup that stops halfway.

1. Update the graphics driver first, since processing leans heavily on the graphics card.
2. Disable real-time protection until the install has completed.
3. Extract the archive to a short folder path and run setup as administrator.
4. Allow the model files to finish installing before launching, as they make up most of the package.
5. Set the output folder and default model in preferences before running a batch.

System requirements and file details

Frequently asked about this title