Star removal is the astrophotography technique that lets you process the sky's structure without wrecking the stars. Here's why it works, an honest look at the tools that do it, and how to remove stars from a Milky Way photo without losing them forever.
Stars sit on top of everything else in the frame. Any move that makes the Milky Way pop (contrast, saturation, clarity, sharpening) also hits the stars, and stars respond badly: they bloat, grow halos, and oversaturate until the sky looks crunchy. Take the stars out first and you can work the sky's real structure (dust lanes, the bright core, airglow bands) as aggressively as the image wants, then return the stars at their original size and color.
It helps to separate two goals that often get lumped together:
Landscape astrophotography adds one more wrinkle: your frame has a foreground. Star removal models were mostly trained on deep-sky images that are all sky, and pointed at a nightscape they will happily chew on foreground highlights too: distant lights, wet rock, snow, specular glints. The ground must be protected, either by masking it yourself before you run the tool, or by using a tool that knows where the sky ends.
StarNet++ pioneered neural-network star removal and it remains free. It runs as a command-line tool or a plugin, cleanly separates stars from nebulosity, and for anyone who needs a starless image now and then it's a proven, no-cost option we're happy to recommend. Its roots are deep-sky, so on a nightscape you handle the foreground protection yourself, and getting the stars back means building a subtraction workflow by hand (more on that below). Our full StarNet++ comparison is here.
RC Astro's StarXTerminator is the polished commercial take: $49.95 one time, excellent results, fast, actively developed, and it runs as a Photoshop, PixInsight, or Siril plugin. It can also generate a stars-only image alongside the starless one, which saves you the subtraction math. If you live in a deep-sky pipeline, it's a superb tool, and we say so plainly in our comparison.
What neither tool solves is the workflow around the result. The output typically comes back as a flattened image: the stars are gone, full stop, unless you build your own star-mask workflow (duplicate the original, subtract the starless version, blend the difference back with Screen or Linear Dodge). That's a well-known recipe and it works, but it's a recipe you assemble by hand every time, plus the round-trip itself if you're exporting to a command line and reimporting. And neither tool offers a middle path: they remove all the stars or none of them.
Before neural networks, the standard Photoshop trick was a Minimum filter or Dust & Scratches applied through a star selection: threshold-select the bright points, expand slightly, then shrink or median them away. It still gets suggested, so it's worth being honest about why it disappoints. These are neighborhood operations that don't know what a star is. They erode fine sky detail along with the stars, leave dark rings and soft holes around anything bright, and fall apart completely in dense Milky Way fields where faint stars overlap real structure. Fine for taming a handful of distracting stars; not a way to produce a genuinely starless sky.
NightScape Lab's Star Removal lives on the panel's Stars tab and was built for nightscapes from the start. Run it on the clean, scene-linear base image and it delivers two ordinary Photoshop layers: a starless layer, and a separate Star Map layer set to Linear Dodge. Stacked together they reproduce your original image; that transparency is the point. The stars are never gone, they're an editable ingredient sitting in your layer stack. The panel's AI sky mask protects the foreground automatically, so distant porch lights and wet granite survive untouched, and there's no export, no command line, no flattened round-trip.
For the many images where you want fewer stars rather than none, the Stars tab also offers Declutter: a flux-sculpted reduction that dims faint clutter in proportion to its prominence, while Constellation Guard protects the bright, round stars so Orion still looks like Orion. It's one slider's worth of decision instead of a removal-and-recombine project, and for a typical Milky Way nightscape it's often the better answer.
Fair to the field: StarNet++ is free and capable, and StarXTerminator is excellent and runs in PixInsight and Siril too. If you're processing deep-sky data in those pipelines, they're the right tools. NightScape Lab's case is the workflow: nightscape-aware removal, a Star Map you keep, and Declutter as the middle path, all inside Photoshop on macOS & Windows.
Because the Star Map is a real layer on Linear Dodge, blending the stars back is ordinary Photoshop work:
In the larger finishing order, star work comes after White Balance and Defringe and before the Foreground Sky Merge, so both plates arrive at the merge already resolved. The full finishing sequence is laid out here.
If you want star removal, Declutter, and the rest of the nightscape workflow (stacking, White Balance, Defringe, the Foreground Sky Merge, Star Trails) in one tool inside Photoshop on macOS & Windows, that's what NightScape Lab is: a one-time purchase, no subscription.
30-day money-back guarantee. If it's not for you, email for a full refund, no questions asked.