Topaz Photo AI

About this release

Topaz built three separate applications for noise, sharpness and size, and then discovered that most photographs need some combination of all three and that deciding which to apply first is its own small expertise. Photo AI is the answer: it analyses the image, works out what is actually wrong with it, applies the right models in the right order, and shows the result with every decision still adjustable.

Noise reduction is the strongest part. Images shot at high sensitivity, in poor light or on a small sensor come back clean while keeping the fine texture that ordinary noise reduction turns into a smooth plastic surface. Raw files are handled before demosaicing, which produces a noticeably better result than cleaning a rendered image afterwards.

Sharpening addresses the causes rather than raising local contrast. Separate models handle camera shake, subject motion and a missed focus plane, and each reconstructs what the detail should have been rather than drawing halos around edges. Faces are recognised and given their own treatment so a portrait recovers eyes and skin texture without turning waxy.

Enlargement uses the same reconstruction approach as the dedicated upscaling application, so a crop that was too small to print becomes usable at size. Everything runs as a plugin inside an image editor or as a standalone application, batch processing covers a whole shoot in one configured run, and output keeps the raw editing latitude where the source allowed it.

What it does

Automatic analysis

The image is assessed and the right corrections are applied in the right order.

Raw noise reduction

Noise handled before demosaicing for a noticeably cleaner result.

Cause specific sharpening

Separate models for camera shake, subject motion and missed focus.

Face recovery

Portraits keep eye and skin detail rather than turning waxy.

Detail preserving enlargement

The same reconstruction approach used by the dedicated upscaling tool.

Editor plugin and batch

Runs inside an image editor or over a whole shoot in one configured pass.

Changed in this version

• New noise model with better retention of fine texture.
• Faster processing on current graphics cards.
• Improved subject detection on cluttered scenes.
• Batch queue handles very large shoots without slowing down.
• Additional raw formats supported on import.

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 models unpack fully before launching.
4. Open the application, process one image, and confirm the graphics card is being used.

Worth knowing before you start

File details