Adobe Photoshop 2026
The complete 2026 raster editor with the full generative model set switched on.
A dedicated image upscaler that enlarges photographs several times over while rebuilding the detail a plain resize would smear, using trained models tuned for faces, landscapes, and low resolution originals.
Overview
Gigapixel does one thing and does it better than a general editor's resize command: it makes a small image large without turning it to mush. A conventional enlargement simply spreads the existing pixels, so the result is soft. Gigapixel instead uses trained models to infer the detail that a higher resolution capture would have held, rebuilding edges, texture, and structure as it scales.
The gain is most obvious on the hard cases. An old scan, a heavily cropped frame, or a low resolution export can be pushed to several times its original size and come out looking like it was shot larger, with hair, foliage, and fabric reconstructed rather than blurred. For print work from small originals, that is the difference between usable and not.
The model selection is what keeps the results honest. Separate models are tuned for faces, for landscapes and general photography, for compressed low quality sources, and for line art, so the program applies the reconstruction appropriate to the content instead of one blunt filter. Face recovery in particular rebuilds features that a generic upscaler would invent badly.
Batch processing handles a whole folder in one run, and the Pro build enables the full model set and the largest output sizes from first launch without a trial limit.
Feature set
Rebuilds edges and texture while enlarging rather than spreading existing pixels into softness.
Separate models for faces, general photography, low quality sources, and line art.
Reconstructs facial detail on small or soft originals instead of inventing it crudely.
Push an image to several times its original resolution for print from small sources.
Enlarge a whole folder in one run with a chosen model and output size.
Runs on its own or as a filter called from a host editor.
Changes in this build
Installation
Hardware
| Processor | Intel or AMD, 64-bit, multi-core |
|---|---|
| Memory | 8 GB minimum, 16 GB for large images |
| Graphics | 4 GB VRAM recommended for accelerated processing |
| Storage | 4 GB install plus output space |
| Display | 1920 x 1080 |
| System | Windows 10 version 22H2 or Windows 11 |
Package
| Title | Topaz Gigapixel AI Pro |
|---|---|
| Version | 8.4.4 |
| Developer | Topaz Labs |
| Category | Photo Editing |
| Platform | Windows 10 / 11 (64-bit) |
| Package size | 1.48 GB |
| Interface languages | English |
| Release identifier | BS-84040001 |
| Listed | 1 year ago |
| Last rebuilt | 4 days ago |
| Total downloads | 581,835 |
Questions
Several times the original in each dimension. The practical limit depends on how much real detail the source held.
No, but one speeds processing considerably. On the processor alone it works, more slowly.
Yes, as a plugin filter from a host that supports external filters, as well as standalone.
Select the face model for portraits. The general model does not reconstruct facial detail as convincingly.
The complete 2026 raster editor with the full generative model set switched on.
The desktop catalogue application photographers actually use for volume work.
The last standalone perpetual release of the photo catalogue and develop application.
A single application that combines noise removal, sharpening, and upscaling into one automatic pass, reading an image to decide what it needs and rebuilding detail rather than smoothing over the problems.