Removing the stars from an image has become a core part of my processing workflow, so when I started a 30-day trial of StarXTerminator I decided to put it properly through its paces against the tool I'd been using all along, StarNet2. I should declare my motivation up front: I've been using the other RC Astro plugins for a while with great success and I'm considering switching, so this comparison is partly me deciding whether to part with my money. I figured I might as well share the process in case it helps anyone else make the same call.
For anyone unfamiliar, the two PixInsight processes on trial are StarNet2, the latest iteration of the free StarNet offering, and StarXTerminator, the Russell Croman (RC Astro) plugin. Both do roughly the same job in slightly different ways: they remove the stars from an image so you can process the starless version independently and screen the stars back in later, or discard them entirely.
I ran both processes on two images, the Whirlpool Galaxy and the Horsehead Nebula, and judged them on four criteria: the quality of the resulting images, the speed at which they run, ease of use, and price. For StarXTerminator I used the generate star image option with unscreen stars enabled, which the tool itself recommends if you plan to merge the stars back in, and I matched the settings as closely as I could in StarNet2.
Round one: the Whirlpool Galaxy
StarXTerminator ran a lot faster than I expected and did a really thorough job. A couple of the bigger stars left slight traces, and zooming right in there's maybe a tiny bit of blue galaxy detail in there, but very few artefacts overall. StarNet2 was noticeably, drastically slower, and while the result also looked pretty good, it left the image a bit blotchy and patchy where the brighter stars had been. Side by side, those artefacts simply don't exist in the StarXTerminator version.
I then applied the same small CurvesTransformation to both starless images (a touch of brightness and saturation) and screened the stars back in using PixelMath. StarXTerminator helpfully displays the screening formula in its dialogue, although annoyingly you can't copy it and there's no help menu, so you have to memorise it. As it happens, it's much the same formula I'd been using anyway. After recombining there wasn't much between them, except a very subtle haze on the StarNet2 image, as though it had lost a little clarity, that the StarXTerminator version didn't have. Bear in mind that artefacts matter more once you start boosting brightness or saturation, because you're boosting the blemishes along with everything else unless you mask them out.
Round two: the Horsehead Nebula
The Horsehead is a tougher test: far more stars, some very big ones, and star halos to contend with. StarNet2 was even slower than before given the extra stars, and the result was peppered with little blotches where stars had been removed, like dried water droplets, particularly around the right hand edge of the nebula. StarXTerminator was again significantly quicker and the image was a lot cleaner; outside the big remaining stars, which were left crisp, everything small was eradicated entirely with no indication stars had ever been there.
Repeating the curves and PixelMath recombination test, the same very subtle haze appeared on the StarNet2 version. I also spotted one star that stayed really crisp in the centre in the StarXTerminator image but had its definition blurred out around the edges in the StarNet2 one. Subtle differences, relatively minor, but noticeable.
The scorecard
On quality, StarXTerminator takes it: fewer blemishes and artefacts, a better starless image, and a better result after screening the stars back in. StarNet2 is still very effective, just not quite as good in my opinion.
On speed, StarXTerminator wins comfortably. On my Apple M1 Mac it took 7 and 10 seconds for the Whirlpool and Horsehead respectively, against 32 and 104 seconds for StarNet2, so four to over ten times slower. Your mileage will vary with the image, your hardware and whatever else is running.
Ease of use I'm calling a tie. Both are very straightforward with barely any configuration, though StarXTerminator gets a nod for showing you the screening formula.
Price, obviously, goes to StarNet2. It's completely free, while StarXTerminator costs just shy of $60, or around $50 if you already own another RC Astro plugin and get the discount.
My verdict
That makes it 2-1 to StarXTerminator ignoring the ease of use tie, or 3-2 including it. Either way, I'll be using StarXTerminator going forwards and upgrading my trial to a full licence when it expires. The quality and speed advantages are worth the price to me, but if the budget is tight, StarNet2 is still a very capable free option. If you want to judge the results for yourself, all of the side by side comparisons are in the video at the top of this post.