StarNet Software

AI-powered software for star removal and noise reduction in astrophotography.

PixInsight Modules

This page describes the StarNet and DeepSNR PixInsight modules: which repository to add, current releases and system requirements, and where the shared installation instructions live.

PixInsight module releases

Supported PixInsight versions: 1.9.4, 1.9.5, and 1.9.6 or newer. Intel Mac modules remain on PixInsight 1.9.4.

StarNet and DeepSNR are now installed in PixInsight through tool-specific repositories. Current releases use a single repository for each tool.

The old TensorFlow/runtime repository path is deprecated for new installs.

Release matrix

Latest available packages for each platform. Windows, Linux and Apple Silicon Mac require PixInsight 1.9.6 or newer. Intel Mac packages require PixInsight 1.9.4.

PixInsight selects the package for its version and architecture. No separate runtime repository is needed.

Current StarNet2 PixInsight release matrix

Operating system Architecture Version / build Backend Release date
Linux x64 2.6.1-0240 ONNX/ORT* 2026-10-30
Windows x64 2.6.1-0240 ONNX/ORT* 2026-10-30
macOS x64 2.6.0-0235 ONNX/ORT** 2026-09-13
macOS ARM64 2.6.1-0240 CoreML 2026-10-30

* Provided by PixInsight.
** Included in our distributions.

Current DeepSNR PixInsight release matrix

Operating system Architecture Version / build Backend Release date
Linux x64 1.3.1-0146 ONNX/ORT* 2026-10-30
Windows x64 1.3.1-0146 ONNX/ORT* 2026-10-30
macOS x64 1.3.0-0141 ONNX/ORT** 2026-09-13
macOS ARM64 1.3.1-0146 CoreML 2026-10-30

* Provided by PixInsight.
** Included in our distributions.

Earlier packages remain available for PixInsight 1.9.4 and 1.9.5. Intel Mac support remains on PixInsight 1.9.4.

The architecture column above refers to the installed PixInsight architecture, not hardware architecture. PixInsight x64 installations on Apple Silicon Macs do not get GPU acceleration even when the hardware has an Apple GPU. For maximum StarNet2 and DeepSNR performance on Apple Silicon, install a supported native ARM64 build of PixInsight.

System and build information

Runtime requirements depend on the PixInsight version listed above. The table below separates what users need to run the package from the systems used to build and validate the release. Your PixInsight version may require a newer operating system.

Current PixInsight module system information

Target Runtime requirement Built and validated on
Linux x64 Linux x64 PixInsight on a compatible glibc-based Linux distribution. Ubuntu 24.04 LTS / WSL2; Linux PixInsight 1.9.4, 1.9.5 and 1.9.6.
Windows x64 Windows x64 PixInsight on 64-bit Windows 10 or 11. PixInsight 1.9.5 and newer use CUDA on compatible NVIDIA setups or CPU; its runtime does not support DirectML. For PixInsight 1.9.4, DirectML runtime libraries are not bundled with StarNet Software packages. They should already be present on most supported Windows installations; missing DirectML components are the exception. See Microsoft's DirectML version history and official DirectML package. Windows 11 x64; Windows x64 PixInsight for validation.
macOS x64 macOS x64 PixInsight on macOS 13.4 or newer. Intel Mac, or Apple Silicon through Rosetta 2. macOS 26.5 Tahoe on Apple Silicon; macOS x64 PixInsight for validation.
macOS ARM64 macOS ARM64 PixInsight on Apple Silicon with macOS 13.1 or newer. macOS 26.5 Tahoe on Apple Silicon; macOS ARM64 PixInsight 1.9.4, 1.9.5 and 1.9.6 for validation.

Current PixInsight module packages do not require TensorFlow, Torch, separate ONNX/ORT installation, or the old /tensorflow/ repository URLs.

Updates in the latest releases

StarNet2 2.6.1

  • Updated for PixInsight 1.9.6 or newer.
  • Windows and Linux use ONNX/ORT provided by PixInsight; Apple Silicon Mac uses CoreML.

DeepSNR 1.3.1

  • Updated for PixInsight 1.9.6 or newer.
  • Windows and Linux use ONNX/ORT provided by PixInsight; Apple Silicon Mac uses CoreML.

Both releases also include minor bug fixes and improvements.

See the StarNet module reference for clipped-pixel restoration and output behavior, and the DeepSNR module reference for settings and input guidance.

Runtime support

  • The latest packages use ONNX/ORT on Windows and Linux, and CoreML on Apple Silicon Mac.
  • PixInsight 1.9.4 Windows modules support DirectML with CPU fallback. PixInsight 1.9.5 and newer Windows/Linux modules use the host runtime: compatible NVIDIA CUDA or CPU, without DirectML.
  • TensorFlow packages are discontinued. Remove old /tensorflow/ repository URLs when updating your repository list.

Installation information