Overview
Traditional noise reduction cannot tell noise from detail, so it removes both and the result looks like wax. This tool was trained to tell the difference, which means a frame shot at a high sensitivity can lose its grain and keep its eyelashes, feather edges and fabric weave. On astrophotography, indoor sport and low light event work the difference between this and a general slider is roughly two usable stops.
Four processing modes cover different material. Standard handles most photographs, clear is aimed at heavier noise where detail preservation matters more, low light is tuned for very high sensitivity frames, and severe noise handles the worst cases where a frame would otherwise be thrown away. A comparison view shows all four against the original at once, which is faster than guessing.
It works on raw files directly, which is the right place to do noise reduction because the data has not been through demosaicing yet. Output can go to a linear raw file so the rest of the editing pipeline still treats it as raw. Batch processing runs a whole shoot with one setting, and the plug-in mode drops it into an existing workflow as a single step inside a host editor.
What it does well
Detail preserving
Noise is separated from real detail rather than averaged away with it, so texture survives the pass.
Four modes
Standard, clear, low light and severe noise, shown side by side against the original for comparison.
Raw processing
Works on raw files before demosaicing and can output a linear raw file so the pipeline continues normally.
Batch processing
Apply one setting across an entire shoot, using the graphics card, with output naming and format control.
Standalone and plug-in
Runs on its own or inside a host editor as one step in an existing workflow.
Changes in this build
- Model set updated with better handling of chroma noise on very high sensitivity frames.
- Processing speed improved on current graphics hardware.
- Raw file support extended to more recent camera bodies.
- Batch queue reworked to keep the interface responsive while it runs.
- Fixed a colour shift when returning results into a host document in a wide gamut space.
What is in the package
- DeNoise installer, 64-bit
- Complete processing model set for offline use
- Host editor plug-in module
- Batch processing queue
- Raw profiles for current camera bodies
System requirements
| Processor | Intel or AMD 64-bit, quad core recommended |
| Memory | 8 GB minimum, 16 GB for batch work on large files |
| Graphics | GPU with 4 GB VRAM strongly recommended |
| Storage | 4 GB free plus output space |
| Display | 1920 x 1080, colour accurate panel recommended |
| System | Windows 10 or Windows 11, 64-bit |
Installation
- Close any host editor before running the installer.
- Extract the archive and run setup.
- Apply the included configuration step before launching.
- Confirm the plug-in appears in the host editor filter menu.
- Disable the update check in preferences.
Before you start
Run it on the raw file rather than after sharpening, the order matters more than the settings do.
The severe noise mode is aggressive and should be reserved for frames that would otherwise be discarded.
Batch output defaults to the source folder, set a separate destination before starting a large run.
Questions about this title
Does it work on raw files?
Yes, and it can output a linear raw file so the rest of the pipeline still treats it as raw.
Can it batch a whole shoot?
Yes, the queue applies one setting across a folder with graphics card acceleration.
Does it need a connection?
No, the model set is bundled and runs locally.
Is a plug-in mode included?
Yes, it runs inside a host editor as well as standalone.
About this listing
This entry was checked on a clean install of Windows 10 / 11 (64-bit) before it was published, and it is rechecked whenever the package is rebuilt. The figures on this page come from the site index rather than from the publisher, so the download count is what people here have actually pulled.
If a mirror stops answering, use the contact form and it gets replaced. Reports about a specific title are handled faster than general messages because they name the file, the mirror and the point at which the transfer stopped.
Requests for a different version, a different language build or an older release go on the requests page. Older versions of a title are usually still held even when only the current one is listed.