Processing Smart Telescope Images: Seestar, Dwarf and Vespera

A Seestar S50, a Dwarf 3 or a Vespera hands you a finished JPEG on your phone after twenty minutes. So why would you process anything?

Because the finished JPEG is not what the telescope collected. It is a live stack, made on a small computer, under time pressure, with decisions taken conservatively so that the preview on your phone never looks broken. The frames underneath it are better than the preview suggests, and on a good night they are much better.

This is about what is actually left on the table, and how to get at it.

What the built-in stack does and does not do

The live stack is doing real work: aligning each new frame onto the last, rejecting the obviously bad ones, and adding to a running average. What it is not doing is most of what a desktop processing run does.

It cannot use frames it has already thrown away. A live stack decides in real time. A desktop run sees the whole session at once and can grade every frame against the session’s own norm rather than against a fixed threshold.

Its gradient handling is limited. Light pollution gradients change through the night as the target moves. Modelling that properly means looking at the whole session.

Its colour is approximate. Proper colour calibration means matching the stars in your field against a catalogue of measured star colours and fitting the white balance to what those stars actually are. That needs a plate solve and a star catalogue.

Its stretch is conservative. The preview has to look acceptable immediately, so the curve is safe rather than optimal.

None of this is a criticism. Doing those things live, on battery, on a phone-class processor, is not the job those telescopes signed up for.

Getting the frames out

Seestar S50 writes individual sub-exposures as FITS files when you enable them. On the telescope’s storage you will find a folder per target containing the subs alongside the stacked result. Copy the whole folder. We have a detailed guide to the Seestar’s FITS output, including what its headers do and do not record.

Dwarf 3 similarly writes FITS subs per session, plus its own stacked output.

Vespera exports FITS when you ask it to; the default is the processed result only.

In every case the thing you want is the folder of individual frames, not the single stacked file. The stacked file is already averaged and there is much less you can do with it.

What changes when you process them properly

The honest version: it depends on how many frames you have and how good the night was.

On a short session in poor conditions, careful processing of the subs may get you very little over the live stack. The telescope was not the limiting factor; the sky was.

On a long session in decent conditions the difference is usually obvious — more faint structure, a flatter background, and colour that looks like the object rather than like the sensor. A hundred ten-second Seestar subs of a bright nebula carry a lot more than the live preview shows.

The one thing that reliably improves is the background. Live stacks tend to sit on an uneven, slightly coloured sky, and modelling that properly across the whole session is the single biggest visible change.

Doing it without learning a new discipline

The obstacle is not the telescope, it is that desktop astrophotography software assumes you already know the sequence — calibrate, register, integrate, gradient, colour, stretch, denoise — and in what order and with what settings.

Siril is free, excellent and cross-platform, and it will do all of this if you are willing to learn which script matches your data.

Akastroid is what we make: point it at the folder of subs and it measures them, works out which decisions your data needs, and shows you the numbers behind each one. Seestar and Dwarf FITS files open directly, no conversion. It is free to try and nothing is uploaded — the frames stay on your computer.

Download Akastroid · The Seestar FITS guide

If you have no computer

A fair number of smart telescope owners do not own a laptop at all, which is rather the point of the product. We are building a cloud version for exactly that case — there is a waitlist.