Topaz Sharpen AI

About this release

Conventional sharpening increases local contrast at edges, which makes a slightly soft image look crisper and makes a genuinely blurred image look like a blurred image with halos. This works differently. It classifies what actually went wrong in the capture and applies a correction built for that specific failure, which is why it recovers detail that a contrast based sharpener cannot touch.

Three problems are handled separately. Motion blur from camera shake is reversed along the direction of the movement. Focus that landed slightly in front of or behind the subject is corrected by rebuilding the edge structure that the lens never resolved. General softness from diffraction, a weak lens, a filter or a lens adapter is treated as a broad sharpening problem across the frame. Choosing the wrong one produces obviously worse results, so the automatic detection is worth trusting first.

Masking keeps the correction where it belongs. Sharpening applied to an entire frame amplifies noise in the sky and in flat shadow areas, so the brush and the automatic subject mask restrict the effect to the parts of the image that carry detail. The before and after view splits at any point in the frame and stays synchronised at full magnification.

It works either standalone on a folder of files or as a plug in from the common image editors and raw processors, where it returns the result as a layer rather than replacing the original. Batch processing runs a whole shoot through one setting, which is what makes it usable on event and sports work where a proportion of every take is slightly soft.

What it does

Blur type detection

Camera shake, missed focus and general softness are identified and corrected by different models.

Detail recovery rather than contrast

Edge structure is rebuilt instead of local contrast being raised, so results avoid the usual halos.

Masking

Brush and automatic subject masks keep sharpening off skies and flat shadow areas where it would raise noise.

Split preview at full magnification

Compare before and after at any point in the frame with both views synchronised.

Batch processing

Run a folder or a whole shoot through one setting rather than opening files individually.

Plug in and standalone

Works on its own or from the common editors and raw processors, returning the result as its own layer.

Changed in this version

• Models retrained for better behaviour on high resolution sensors.
• Faster processing on current generation graphics cards.
• Automatic subject masking is more accurate around hair and foliage.
• Reduced artefacts on high contrast edges in the focus correction model.
• Batch queue gains pause and resume.

System requirements

Install order

1. Unpack the archive and run the installer.
2. Let the model files finish installing before opening the application.
3. Select the graphics card as the processing device in preferences.
4. Install the plug in bridge from the application if you want it inside an image editor.

Worth knowing before you start

File details