System requirements

What BiggSky needs to run on each platform, and which MATLAB Runtime each build uses.

The short version

If your machine is from the last few years, it will run

BiggSky is a desktop application. Nothing is uploaded and nothing is processed in the cloud — the work happens on your hardware, and more of it helps.

You do not need a MATLAB licence. BiggSky is built on MATLAB, and the MATLAB Runtime it needs is free. On Windows and macOS it installs alongside the application; on Linux you install it yourself first.

Every platform

  • 8 GB RAM minimum
  • 16 GB or more strongly recommended for large SER files
  • Several GB of disk for the MATLAB Runtime
  • Room for your captures — they are far larger than anything BiggSky writes
  • An internet connection to activate; none needed afterwards
Per platform

Operating systems and runtimes

Each build ships against a particular MATLAB Runtime release. The authoritative operating-system list for each is the MathWorks document linked in the last column.

BiggSky builds, supported operating systems and MATLAB Runtime releases
Build Operating system MATLAB Runtime MathWorks requirements
Windows Windows 10 or 11, 64-bit R2023b R2023b Windows (PDF)
Windows — Blackwell GPU Windows 10 or 11, 64-bit, with an NVIDIA RTX 50-series card R2025b R2025b Windows (PDF)
Mac — Apple Silicon macOS on M-series. A native arm64 build, not Rosetta. R2025a R2025a macOS (PDF)
Mac — Intel macOS on Intel, 64-bit R2023b R2023b macOS (PDF)
Linux 64-bit Intel or AMD R2024b R2024b Linux (PDF)
Why the runtime versions differ. Each build is compiled against the runtime that best supports its hardware: the Blackwell build needs a newer one for the GPU, and the Apple Silicon build needs a newer one to run natively. The application is the same; only the runtime underneath it changes. The full list is on the MathWorks previous releases page.
Graphics acceleration

GPU support

Optional everywhere. BiggSky runs perfectly well on the CPU — a GPU makes it faster, it does not unlock features.

GPU acceleration support by build
Build GPU acceleration
Windows — standard NVIDIA cards of Compute Capability 3.5 or higher — Kepler (2013) onwards, which is most cards still in use. It does not support Blackwell. An RTX 50-series card will simply not be used by this build. Check your card’s compute capability.
Windows — Blackwell For RTX 50-series cards. This is the only build that supports them, which is the whole reason it exists.
Linux Not tested. It may work; we have not verified it, so we are not claiming it. If you try it, we would genuinely like to hear how you get on.
macOS No GPU acceleration on Intel or Apple Silicon. Processing runs on the CPU, which on an M-series machine is quick.
GPU acceleration is not included in SSE Lite. It comes with SSE Pro, BiggSky DSO and Universe. Lite processes on the CPU, and produces identical results — it simply takes longer. Compare the editions.
You do not need a fast machine. BiggSky was developed on a 2008 Dell workstation with a GTX 780 Ti and 3 GB of video memory — a card that sits exactly on the Compute Capability 3.5 floor above. If it runs there, it will run on yours.
No NVIDIA card at all? Nothing is lost. Use CPU Multi-Core spreads the work across your processor cores instead, and every feature behaves identically either way.

File formats it reads

8-bit, 16-bit and 32-bit variants across the still formats.

Stills
TIFF · PNG · BMP · GIF · JPEG · Camera RAW · FITS · XISF · DICOM

Video and sequences
SER · AVI · other common movie formats · numbered image sequences

A note on disk and memory

Your captures are the large thing here. A SER or AVI of a lunar or planetary session runs to gigabytes across hundreds or thousands of frames.

BiggSky does not add to that. It writes no scratch files and no calibration cache — processing happens in memory, and what lands on disk at the end is a single image file, small next to the capture it came from. The one thing that can be large is optional and deliberate: you may ask it to save the aligned, warp-corrected frames as a new SER for review.

Bit depth is yours to choose in Save Options: 8, 16 or 32-bit. Internally the data is always held as 32-bit floating point and kept linear, which is what lets the deconvolution work properly — but you decide what gets saved.

Because there is no scratch space, memory matters more than disk. Processing is quickest when the file is under about half your available RAM.

Think it will run?

The surest test is the free one. Version , fully featured for 30 days.