Topaz Photo AI
Version 4.1.0 · Topaz Labs · Windows 10 / 11 (64-bit) · ranked #59 in the library
A single pass image improver that works out what a photograph actually needs, then applies noise reduction, sharpening, enlargement and face recovery in the right order without you tuning six sliders.
Read the installation section below before you start. The order of the steps matters on this release.
About this release
Overview
The premise is that the three problems most photographs have are related and should not be fixed separately. Noise reduction applied before sharpening produces a different result from the same two in reverse, and both interact with any enlargement. This application analyses the image, decides which of its models are needed and in what order, and applies them as one operation with the settings already sensible. Most images need nothing beyond accepting what it suggests.
You can override all of it. Each enabled module has its own strength controls, the model choice per module can be forced, and a subject mask restricts processing to a region so a background stays untouched. The preview is a split view at full resolution with a synchronised zoom, which is the only honest way to judge whether a denoise pass has taken texture away along with the noise.
It works as a standalone application and as a plugin inside the main raster and catalogue applications, which is where most people use it. Send an image across, process, and it returns as a layer or a new version without a manual round trip. Batch mode processes a folder unattended with per image analysis rather than one shared setting, which is what makes it usable on a full shoot rather than a single frame.
What it does
Feature breakdown
Automatic analysis
Detects which corrections an image needs and applies them in the correct order, with the settings already reasonable before you touch anything.
Noise reduction with texture retention
Removes high sensitivity noise while keeping fine detail, with separate handling for raw and already processed files.
Sharpening for the right cause
Distinct models for motion blur, missed focus and general softness, since the three need different treatment.
Face recovery
A dedicated pass over detected faces so a rescued portrait does not come back looking plastic.
Enlargement in the same pass
Increase resolution as part of the same operation rather than as a separate step that fights the denoise result.
Standalone, plugin and batch
Works on its own, as a plugin inside the raster and catalogue applications, and across a folder unattended with per image analysis.
Changed in this build
Version 4.1.0
- New denoise model producing better results on very high sensitivity raw files.
- Subject masking improved on images with fine edges such as hair and foliage.
- Faster processing on cards with limited video memory through improved tiling.
- Batch queue now reports per image which modules were applied, rather than only a completion status.
- Raw handling improved so the application reads the file directly rather than a converted intermediate.
Installation
Follow these in order
- Go offline before installing.
- Extract the archive.
- Disable real time protection.
- Run the installer as administrator and let it register the plugin with any host applications it detects.
- Apply the included licensing step before first launch.
- Open the application, run one image through to confirm the models load, then restore protection.
Release notes
From whoever packed it
- The models are bundled so nothing downloads on first run.
- Block the application in your firewall, an online check reverts it to trial behaviour.
- If the plugin does not appear in a host application, install the host first and then run this installer again.
System requirements
Minimum and recommended
| Operating system | Windows 10 version 22H2 or Windows 11, 64-bit |
|---|---|
| Processor | Modern x64 processor with AVX2 support |
| Memory | 16 GB recommended, 8 GB minimum |
| Graphics | GPU with 6 GB VRAM or more recommended, processing runs on the processor but far more slowly |
| Storage | 6 GB for the application and the bundled model set |
Questions about this release
Answered before you ask
- Is it different from the dedicated upscaler?
- It overlaps. This one decides what an image needs across several corrections. The upscaler does one job with more control over it.
- Does it work on raw files?
- Yes, it reads raw directly, which gives better denoise results than processing an exported file.
- Can I turn the automation off?
- Yes, every module can be enabled, disabled and tuned by hand.
Download
Pick a mirror