Guide
GraXpert or Siril for Seestar images: which does what
People ask which one to learn first. They do different jobs, they are both free, and most people who process Seestar files end up using both. This is what each one is for.
Siril
Siril is a complete program for deep-sky images. It runs on Windows, macOS and Linux and has a tutorial written for the Seestar. On a Seestar file it does these jobs:
- stacks raw subs, with a script made for the Seestar’s files;
- solves the position of the image from its stars (the Seestar writes the coordinates into the FITS header) and calibrates the colour against a star catalogue;
- removes a background gradient, from sample points you place;
- stretches the image, and sets saturation;
- saves the result as FITS, TIFF, PNG or JPG.
Siril is the one you cannot do without, because it is the one that stacks and the one that stretches. It has many tools, and it takes time to learn.
GraXpert
GraXpert is a smaller program, also free and open source. It does three jobs, and it does them with trained models:
- Background extraction, to remove a gradient. It can work with no sample points at all, or with points that you place.
- Denoising, which removes the grain that is left in a short exposure from a small lens. It works best on the linear data, before the stretch.
- Deconvolution, which sharpens stars and fine detail. This guide does not cover it.
It does not stack, and it does not calibrate colour or stretch. You open a file that is already stacked, run the step and save.
Where they overlap
Both remove a gradient. Siril’s method fits a model to the sample points you place on empty sky, so you decide what counts as sky. GraXpert’s trained model needs no points and is quicker to start with. It can be wrong where it takes part of a large object for sky, and so can sample points that are placed badly. The gradient guide covers how to check.
Siril can call GraXpert from inside its own window, so you do not need to switch programs once both are installed.
A Seestar collects light through a small lens, often for under an hour, so the faint end of a stack is mostly noise. This is the reason to add GraXpert’s denoise to a Siril workflow.
The order that works
- Siril: stack the raw subs, or open the stacked file.
- GraXpert or Siril: remove the gradient.
- Siril: plate solve and calibrate the colour.
- GraXpert: remove noise, on the linear data.
- Siril: stretch, then saturation, then save.
Save the files between the steps as FITS, not JPG. Save a JPG only at the end.
Which one first
Learn Siril first. With the gradient removal and the colour calibration in Siril alone you can finish an image. Add GraXpert when noise or a stubborn gradient is the problem you want to solve.
What Stelora does
Stelora runs these two programs for you, in a fixed order: Siril stacks, GraXpert removes the gradient, GraXpert removes the noise, and then the white balance is set from the stars. Siril stretches and adds saturation, and code I wrote corrects the star colour and the local contrast. This is not the order of the list above: it removes the noise before it balances the colour. You get the result without installing either program. It is a fixed recipe, so a patient person who tunes each step in Siril will beat it on a given image. How it works says where.