Topaz Gigapixel AI Pro

About this release

Enlarging an image has always been a lossy compromise. Traditional interpolation has no idea what is in the picture, so it averages between the pixels it has and produces something soft and slightly plastic. Gigapixel works the other way around: models trained on very large sets of real photographs infer what the missing detail should have been, and reconstruct it. Hair, fabric weave, foliage and brick come back as texture rather than as blur.

Several models ship because no single one suits every source. There are models for standard photographs, for low quality compressed images, for portraits where facial detail matters more than anything else, for scanned film with grain that should be respected rather than smoothed, for line art and for computer generated imagery. The comparison view runs a candidate model on a region so you can see the difference before committing to a full pass.

Restoration work is where it earns its place most obviously. Old scans, screen grabs, compressed images pulled off a phone and film frames all clean up in the same pass as the enlargement, since the models handle compression artefacts, noise and softness at the same time as the scaling. A photograph that could not be printed larger than a postcard becomes something that holds at poster size.

The Pro build adds the parts that matter for volume. Batch processing runs a whole folder with one configured recipe, the command line interface fits into an existing pipeline, and the plugin path lets the process run from inside an image editor without exporting and reimporting. Output preserves colour profile and metadata rather than stripping them.

What it does

Detail reconstruction

Texture is inferred and rebuilt rather than interpolated, so enlargements stay sharp.

Model per source type

Separate models for photographs, compressed images, portraits, film scans, line art and rendered imagery.

Face recovery

Facial detail is treated separately so portraits do not turn waxy at large scales.

Artefact and noise handling

Compression damage and grain are addressed in the same pass as the enlargement.

Batch and command line

A whole folder runs from one recipe, and the process fits into an existing pipeline.

Editor plugin

Runs from inside an image editor with no export and reimport step.

Changed in this version

• New model with better behaviour on heavily compressed sources.
• Faster processing on current graphics cards.
• Comparison view gains a synchronised zoom across models.
• Batch queue handles very large folders without slowing down.
• Colour profile and metadata preservation improved on output.

System requirements

Install order

1. Unpack the archive to a folder with room for the model files.
2. Update the graphics driver before installing.
3. Run the installer and let the model files unpack fully.
4. Open the application once and run a single small image to confirm the graphics card is being used.

Worth knowing before you start

File details