A complete imaging session, covering equipment, guiding, plate-solving and the sequencer, running on a Raspberry Pi, a Windows mini PC, or anywhere else, controlled from the browser of your phone, laptop, or warm-room tablet.
N.I.N.A. Polaris is a community-driven fork of N.I.N.A. It is not affiliated with or supported by the official N.I.N.A. team. For anything Polaris, open an issue on this repository; please do not contact the N.I.N.A. developers about it.
From connecting the hardware to a finished image: capture, guiding, framing, live stacking, sequencing, and AI post-processing, all driven from the browser.
Talk to your gear over INDI (400+ Linux drivers), over ASCOM and Alpaca across the LAN, or through native camera SDKs for ZWO, Player One, and Canon, Nikon and Sony DSLR/mirrorless. Auto-discovery finds devices on the network.
A native PHD2 implementation drives guiding directly, with a live RA/Dec chart, dither, settle, and profiles. When you want the original PHD2 interface, fall back to the real application embedded right in the browser over xpra or VNC.
A full sky explorer powered by the Stellarium Web Engine, with total mount integration. Pick any target, slew to it, then plate-solve and center automatically, all from the browser.
EAA-style live stacking with registration, integration, and real-time stretch over a bandwidth-adaptive WebSocket stream, so the image builds up live as each frame lands.
Record high-frame-rate video for planetary, lunar, and solar work, then process it on the spot for lucky-imaging results.
A tree-based sequencer with containers, loop conditions, and event triggers: auto-focus on temperature or HFR drift, meridian flip, dither, safety abort. Drag and drop the whole plan in the browser.
Automated focusing with V-curve and HFR fitting, kicked off by temperature change or HFR drift during a sequence, so stars stay tight all night.
A complete STUDIO for stacking and full post-processing, including AI background extraction, denoise, and deconvolution via GraXpert, accelerated on the client GPU through the browser or on the SBC's NPU.
OpenCL-backed GPU acceleration for the heavy image-processing paths like debayer, stretch, and stacking, with a transparent CPU fallback when no GPU is present.
The built-in benchmark reports a single Polaris score so you can pick a board. Higher is better. Figures below are Release builds.
| Device | Cores | Polaris score | Stacking | Capture | Memory BW |
|---|---|---|---|---|---|
| Raspberry Pi 4 Model B | 4 cores | 22.2 Mpx/s | 11.5 Mpx/s | 4.7 GB/s | |
| Raspberry Pi 5 Model B | 4 cores | 39.9 Mpx/s | 27.3 Mpx/s | 7.7 GB/s | |
| Orange Pi 5 Pro (RK3588S) best value | 8 cores | 46.2 Mpx/s | 30.9 Mpx/s | 23.1 GB/s | |
| Core i9-13900KF + RTX 5070 | 32 threads | 99.4 Mpx/s | 89 Mpx/s | 41 GB/s |
Polaris score is the headline figure; Mpx/s columns are live-stack and capture throughput. The Orange Pi 5 Pro is shown with the OpenCL GPU backend on (242; 227 CPU only) and the Pi 5 at its best, actively cooled run. Full per-device tables and methodology in the benchmark docs.
MIT-style license inherited from upstream N.I.N.A. Bring your own scope.
Free & open-source · MIT-style license inherited from upstream N.I.N.A.