One trail found, one line blanked
Made-up test frame. No real sub has been through it yet.
- 14 of 14test trails found
- 0false trails in 11 clean frames
- 2.1 sper 24 MP sub on our Mac
How we tested it
We made 25 test frames with a fixed seed and ran them headless in PixInsight 1.9.4 on our Mac on Nov 4, 2026.
They hold a faint trail, a bright wide one, a crossing pair, a partial trail and a color frame. Real subs, with nebulae and registration noise, come next.
The line width follows each trail: 15 px for this thin one, 36 px for the bright, saturated one.
Three steps, then stack as usual
- Pick a folder of registered subs: XISF or FITS, mono or color.
- The script finds straight trails and writes a copy of each sub with a thin line of zeros over each trail.
- Stack the copies with a low rejection range. The zeros drop out, so the trail never reaches your master.
- Blanks, never paints. Every pixel outside the line stays bit for bit.
- No neural net, no GPU. Classic image math on the CPU. You can see the mask.
$19 now, $29 after launch
- Ships by Dec 18, 2026. By email, to the address you give Stripe.
- Fewer than 5 by Nov 18? We refund every buyer in full and stop.
- Not happy after delivery? Ask within 14 days for a full refund.
Questions
Why not let rejection remove the trails?
Rejection works well with many subs. With fewer than about 20, or with a bright, wide trail, edges can stay in the master. A blank line gives rejection a clear target.
Which PixInsight versions?
1.9.4 and later. It is a script, so it runs where PixInsight runs. We test on 1.9.4 on a Mac.
Does it change my files?
No. It writes new copies, with the same file names, into a folder you pick. Your originals stay as they are.
What is not in the first version?
Curved trails, mosaics over 60 MP, and any mode that paints over a trail in a finished master.
Who makes it?
Brasspage, a very small web studio. Tack, our AI personality, writes the code and runs the tests. The trail finder itself is classic image math, not a neural net.