Photon Finish is an automated astrophotography processing pipeline for ZWO Seestar S30 and S30 Pro stacked color FITS captures. It turns a Seestar stack into a finished photograph—locally, automatically, and without changing the original capture.
This is a hobby project for Seestar astrophotographers who enjoy taking the photos more than learning a multi-application editing workflow. It coordinates GraXpert, Siril, StarNet2, and ExifTool to remove gradients, reduce noise, reveal faint detail, manage stars, prepare useful close-ups, and export finished images.
The Seestar telescopes and these open-source tools are already remarkable. This project does not replace them: it supplies an automated, measured workflow across them. It would not exist without the work of ZWO and the communities behind those tools.
It may be a good fit if you:
- have a ZWO Seestar S30 or S30 Pro;
- save the telescope's stacked color
.fitor.fitsfile; - want a polished result without learning several photo-editing packages; and
- are comfortable letting the app make conservative choices from the image.
It may not be the right fit if you want complete manual control, want to calibrate and stack raw subframes, or expect software to repair a poor capture. The specialist tools are better places to learn a fully hands-on workflow. Compact and emission nebulae, galaxies, and star clusters are the validated core; very large, frame-filling nebulae remain best-effort.
- Works from a qualifying Seestar stacked color FITS, not the stock JPG.
- Removes uneven light-pollution gradients and reduces noise.
- Corrects orientation, develops color, and stretches faint signal into view.
- Detects satellite trails and attempts a conservative repair where supported.
- Creates useful star treatments and a galaxy close-up when warranted.
- Exports a 16-bit TIFF, PNG, share-ready JPG, metadata, a processing record, and an optional adjustment recipe.
- Checks the finished result and reports when the capture limits what can be recovered.
- Processes locally. Source captures are treated as read-only.
The app does not control the telescope, capture images, stack raw subframes, upload your photographs, or invent detail that the data does not contain.
The v1.0 release candidate supports:
- macOS: the release build is a signed, notarized app distributed in a DMG.
- Windows 11: a packaged application is coming soon. Run from source in the meantime.
Developers on either platform can also run from source.
All installations require GraXpert, Siril, StarNet2 (including its command-line executable), and ExifTool to be installed separately.
Read:
For a quick dependency check after a source installation:
seestar doctorPackaged Mac users do not need seestar on PATH; the app checks its helper
tools and reports anything missing.
- Your originals remain yours. Choose a separate output folder; the app never edits or deletes the Seestar source library.
- Your images stay on your computer. The optional SIMBAD lookup sends only sky coordinates, never image data. A bundled Messier and Caldwell catalog supports many targets offline.
- The photograph is the authority. Catalog identity may help classify an uncertain field, but measurements and the FITS filter remain authoritative for processing.
- Uncertainty is reported. A qualified result is better than false confidence.
See Privacy and network use and Capabilities and limits.
The Seestar app now includes its own post-processing, and it keeps improving. Its AI Enhance and AI Denoise tools give a one-tap, on-device result that is genuinely convenient—if that meets your needs, use it. This project exists for a different goal, and the two are not interchangeable.
- Works from the stacked FITS with dedicated tools. It processes the linear 16-bit master through GraXpert, Siril, and StarNet, preserving the full dynamic range and giving you inspectable, professional-grade control over every stage.
- Measured, never generative. Some AI enhancement can add "detail" the data does not contain. This project only reveals what the photons support, and reports when a capture cannot be recovered—it never invents structure.
- Transparent and repeatable. Every result ships with a processing record and an adjustment recipe you can read, tune, and re-run. In-app sliders leave no comparable trail.
- Standards-grade output. A 16-bit TIFF, a lossless PNG, a share-ready JPG, and embedded capture metadata—not only an enhanced JPG.
- Offline and non-destructive by design. Processing runs on your own computer, and your source library is never modified (see the promises above).
If you want the fastest possible "good enough" image with no extra software, the Seestar app is the right tool. If you want a faithful, inspectable result that gets the most out of the same capture, that is what this project is for.
The engine is a deterministic functional core with an imperative shell around the external tools. Start with:
The detailed operator and engine contract remains in
AGENTS.md — the vendor-neutral entry point read by Codex,
Cursor, and other AI coding tools. CLAUDE.md is a thin pointer
into the same content for Claude Code. Historical design records remain in
the private maintainer repository and are not part of the public project.
Photon Finish is the public project name. The Python distribution is
photon-finish; for v1 compatibility, its import package and command-line
interface remain seestar, and the macOS application is Photon Finish.app.
This repository is preparing the v1.0 release. See CHANGELOG.md for release contents and SUPPORT.md for help.
The application is MIT licensed. Its bundled known-target catalog is derived from OpenNGC and distributed under CC-BY-SA-4.0. External applications are installed separately and keep their own licenses. See THIRD_PARTY_NOTICES.md.
This independent hobby project is not affiliated with or endorsed by ZWO, GraXpert, Siril, StarNet, or ExifTool.