Guide

How to calibrate, stack and finish a night's imaging in AstrylStudio — written for someone opening it the first time.

Every menu path and button name here is the one in the current build. Screenshots and recorded walkthroughs follow.

AstrylStudio is a Windows desktop application. Everything on this page happens on your own machine: frames are read from your disk and written back to it, and no image is uploaded anywhere.

During the private beta you need an account to open the app. That account is also what carries your plan; it is not a licence dongle and the app keeps working while your machine is offline.

Drag FITS, XISF, PNG, JPEG or TIFF files onto the workspace, or use File > Open Image. Several images can be open at once, each in its own window.

A linear astronomical frame looks black until it is stretched. Press Ctrl+A or F12, or right-click the image and choose STF Auto-Stretch — this changes what you see, not the pixels, and the same keys put it back to Direct View. Saving after an auto-stretch still writes the original linear data; Quick Auto HT to Data in that menu is the one that writes the stretch into the pixels.

Tools > Data Library reads a folder tree once and builds an index of every frame in it — target, filter, exposure, binning, date — so the rest of the app can ask questions instead of you opening files to find out. Build / Refresh Index is the only step that reads everything; after that it is a database.

Frames arrive with the target spelled several ways across nights, and the folder often disagrees with the OBJECT header. The library shows both columns side by side, and where they disagree you can Ask the sky — it plate-solves the frame and tells you what is actually in front of the telescope, which settles it without you guessing.

Preview Plan shows every move it would make and changes nothing. Apply Plan… carries it out and writes a journal, so the whole run can be undone afterwards. Copy keeps the originals; move on the same drive is instant.

Script > Auto Integrate Script (AIS) is the whole path from raw frames to a finished stack. It groups frames by filter, binning and exposure from the headers, so a mixed folder does not need sorting by hand, and one-shot-colour frames are debayered automatically from the BAYERPAT header — mono frames are untouched by it.

  1. Load your frames. Scan Folder auto-classifies a folder tree by target and date; Library takes a target straight off the shelf if you built an index in step 03. Otherwise add BIAS, DARK, FLAT and LIGHT by hand.
  2. Check the groups it found. A filter or exposure you did not expect almost always means a header is wrong, and it is much cheaper to fix there than downstream.
  3. Decide how far to go. Preprocessing Only stops after calibration; leaving Enable Registration and Enable Integration on takes it through to a stack in one run.
  4. Run it. Master frames land in masters/, calibrated lights in light/ with a pp_ prefix, and the stack in the run folder.

Three switches are worth knowing before the first long run. Large-Scale Rejection removes satellite and aircraft trails and passing cloud, which ordinary pixel rejection is too local to catch. Split by exposure keeps 120 s and 240 s subs in separate stacks instead of averaging two different noise regimes together. Auto-remove black borders crops the ragged edge registration leaves behind. Resume skips filters that already finished, so an interrupted run does not start over.

Every run writes its own reasoning down, and the buttons to read it are in the same window.

  • Open Quality Plans — one CSV per filter, a row per frame: measured FWHM, eccentricity, background, noise, the score, and the weight it was given. This is where you look when a stack disappoints; it usually names the night that did it.
  • Open Manifest — run_manifest.json, what was run with which settings.
  • Open Run Folder — the output itself.

Nothing here needs the app to read it. A CSV opens in a spreadsheet, and the stack is an ordinary FITS.

The processes along the top bar apply to the active image window. Drag a process icon onto an image to run it there, or open it and use Preview before Apply — every destructive process has both, and Preview never touches your pixels.

Start here: Background Gradient Remover, then colour calibration, then deconvolution, then a permanent stretch with Histogram Transformation, Curves or Generalised Hyperbolic Stretch, then noise reduction, sharpening and saturation. Everything up to the stretch happens on linear data.

That is a working default, not a law — and it is worth knowing which parts are which, because you will read confident advice that contradicts it.

Settled

Deconvolution runs on linear data. Every source agrees, and the app warns you if you try it on stretched data.

Stated by the people who wrote the tools, with nobody arguing back

Colour calibration — ordinary or photometric — also wants linear data. Once photometric calibration has set your colour, changing the balance afterwards discards what it measured. Star removal after a heavy stretch tends to fail.

Genuinely unsettled

Where noise reduction goes, whether it belongs before or after deconvolution, where star removal belongs, and what to do about a green cast are all argued over by people who know what they are doing — and the published guidance contradicts itself. One vendor writes that noise reduction must not precede deconvolution; another program's official documentation says do it first. The green-cast tool has four camps, one of which is that you should not use it at all. If your result looks right and you got there in a different order, you are not doing it wrong.

What to look at before moving on

AfterLook atMove on when
Background removalThe background, stretched hardIt is flat, and the target has not lost its faint outskirts
Colour calibrationStar coloursThey span white through orange, and the sky is neutral
DeconvolutionBright stars at 1:1They are tighter, with no dark ring around them
The stretchThe histogram's left edgeNothing is clipped to black
Noise reductionFine grain at 1:1, then the whole frameThe grain is softer and no structure has turned plastic

Linear and stretched are not interchangeable, and getting them the wrong way round is the commonest way a first night goes wrong. The app tells you which one it thinks the image is, and warns before a process that only makes sense on linear data. If it has guessed wrong, the image right-click menu lets you say so.

Solving needs a local Gaia catalogue. Tools > Star Catalog (Gaia) downloads one for the sky you actually shoot: it asks the archive for the field in front of it and merges the result into a single file, so the catalogue grows to match your targets instead of starting at tens of gigabytes.

Before an observing trip with no network, put your target list into the prefetch box in the same window. Targets already covered cost nothing, so the same list can be re-run before every trip.

File > Save writes FITS: linear pixels, full precision, the working format.

File > Export Image (Ctrl+Shift+E) is the one to send someone. It writes TIFF, PNG or JPEG with the stretch you are looking at baked into the pixels — a FITS saved after an auto-stretch still holds linear data and opens nearly black anywhere else.