One click, against the alternatives

At a glance

What one run of each tool gives you, starting from the frames the Origin saves. Every one of these tools can produce a fine image in skilled hands; the question here is how much of the work is done for you.

Features of the other tools as described in their own documentation, October 2026. If something here is out of date, tell us and it will be corrected.

	OriginStack	Siril	DeepSkyStacker	PixInsight	Astro Pixel Processor	Origin app
Raw frames to a finished image in one run	Yes, one click	Stacking by script; finishing is a series of separate tools	Stack, then basic levels and saturation	Stacking automated (WBPP); finishing is a chain of processes you run one by one	A tab-by-tab workflow; finishing tools run one by one	Yes, on the telescope
Reads the Origin session folder	Target, sky position, camera gain and calibration frames	As generic FITS	As generic FITS	As generic FITS	As generic FITS	It writes it
Picks settings for the target	Recognises galaxy, nebula, cluster, star field, comet	You choose	You choose	You choose	You choose	Not stated
Gradient removal	Automatic	Separate tool	No	Separate process	Separate tool	Automatic
Colour calibrated against Gaia stars	Automatic	Separate tool	No	Separate process	A colour calibration tool	Not stated
Noise reduction, star reduction, local contrast, stretch	Automatic, tuned per target	Separate tools	Levels only	Separate processes, some as paid add-ons	Separate tools	Not stated
Learned deconvolution for galaxies	Automatic on Origin data	Classical deconvolution tool	No	Classical, or a paid add-on	Not stated	Not stated
Keeps the linear stack for further work	Yes, a FITS file	Yes	Yes	Yes	Yes	A TIFF with a large pedestal
Price	Free, open source	Free, open source	Free, open source	Paid licence	Paid licence	Comes with the telescope
Runs on	Windows, macOS, Linux	Windows, macOS, Linux	Windows	Windows, macOS, Linux	Windows, macOS, Linux	The telescope, via a phone or tablet

Astro Pixel Processor is in the table for its features; it is also measured below, on one session.

What the one click does after the stack

Stacking is where most tools stop. In OriginStack it is the halfway point: the finishing chain below runs in the same click, in this order, with settings chosen for the kind of target it recognised. Each step can be switched off, and every one is described in how it works.

Colour calibrated against Gaiaon by default

Before finishing starts, the stack's colours are set from the stars in the field and their catalogued Gaia colours, so a Sun-like star comes out white and the rest follow. No reference image or manual white balance.

Hot pixels and star maskon by default

Leftover hot pixels are cleaned per colour channel, and the stars are found once and reused, so every later step knows what is a star and what is sky or nebula.

Gradients removedon by default

Light pollution, moonlight and vignetting leave a slope of brightness across the frame. A background model is fitted to the sky between the stars and the object, each colour against its own sky, and subtracted, with the object protected.

Colour noise and a neutral skyon by default

Coloured speckle in the dark sky is smoothed without touching the brightness detail, and the sky is set to a neutral, even black level instead of a tinted one.

Noise reductionon by default

A wavelet denoiser that protects elongated structure such as filaments and spiral arms, chosen after benchmarking the alternatives on scenes with known truth.

Learned deconvolutionautomatic for Origin sessions

A small neural network trained on real Origin sessions, against sharper survey images of the same sky, undoes the night's seeing blur and removes noise in one step. On held-out targets it gave the narrowest stars of every option, with no dark rings and no invented detail, checked against independent stacks.

Star reduction and local contraston by default

Stars are narrowed without losing their colour or peak, so they stop crowding out the object, and mid-scale detail on the object is lifted against the noise at each scale.

Edges and the final skyon by default

The ragged border left by alignment is cleaned up, and the sky is flattened and neutralised one last time after everything else.

The stretchon by default

A generalised hyperbolic stretch brings up faint light while keeping bright cores from blowing out, and an extended bright region is rolled off rather than clipped. You get the JPEG, and can pick another of ten stretches afterwards without running again.

Measured, not claimed

The same frames and the same bias, dark and flat for every tool, on one machine, each at default-style settings. Star width is fitted on the same stars in each pair of stacks, each on its own pixel grid. The full method, and the times these numbers were corrected, are on the home page.

Star width on the same stars, OriginStack against each tool (smaller is sharper).

Session	vs Siril	vs DeepSkyStacker	vs Astro Pixel Processor	vs StackingWizard	vs the Origin's own stack
Omega Nebula, 114 frames	4.42 / 4.56 px	4.42 / 4.83 px	2% narrower	4.39 / 4.61 px	–
Sunflower Galaxy, 158 frames	3.51 / 3.60 px	–	–	–	3.54 / 4.25 px
Sculptor Galaxy, 532 frames	2.66 / 3.01 px	–	–	–	–
Crab Nebula, 133 frames	2.54 / 2.77 px	–	–	–	2.54 / 3.50 px
Black Eye Galaxy, 81 frames	8.21 / 8.26 px	–	–	–	8.15 / 8.38 px

A handful of sessions from one camera are a small sample, and the other tools have many settings that were left at their defaults. tools/bench_vs_siril.py repeats the comparison on your own frames and scores a stack from any other program with --external; tools/bench_vs_origin.py does the same against the Origin's own stack.

Fast, and the finished image included

The Omega Nebula, 114 frames, on an 8-core laptop: 62 seconds from the raw files to the finished image. To the stack alone, OriginStack is faster than Siril on all three sessions measured, even though it also scores, white-balances and finds stars in every frame.

Time to the stack, OriginStack against Siril, run back to back on the same machine; and OriginStack's time to the finished image.

Session	OriginStack, to the stack	Siril, to the stack	OriginStack, finished image
Omega, 114 frames	44 s	54 s	62 s
Sunflower, 158 frames	54 s	67 s	81 s
Sculptor, 532 frames	128 s	159 s	144 s

Astro Pixel Processor (2.0.0-beta48, defaults) took 3 minutes 47 seconds on the Omega frames, timed by hand in its window: 23 s to calibrate, 58 s to register, 40 s to normalise and 105 s to integrate. In the run measured against it OriginStack took 57 seconds to the stack and Siril 66, so its times are not comparable to the table above, which comes from an earlier run; APP was run once, on this one session.

StackingWizard (beta, build 2026-10-31, defaults) took 3 minutes 2 seconds on the same frames, timed by hand: 22.5 s to grade the lights and 2 min 39 s to stack. It kept 80 of the 114 frames and demosaics with bilinear interpolation. In the run it was scored against, OriginStack took 55 seconds to the stack. One session, one run.

DeepSkyStacker took 14 minutes 8 seconds on the Omega frames at its defaults, in an earlier run.

Over 80 kernels in Rustbuilt in

The hot paths are compiled native code: calibration, debayering, hot-pixel repair, alignment warps, the stacking combines, background fitting, star detection, denoising and photometry. Each has a pure-Python fallback, and most return exactly the same numbers as the code they replaced.

Measured against the NumPy and SciPy code they replaced: the default sigma-clip stack about 37 times faster, the fused weighted combine about 100 times, hot-pixel repair 25 to 55 times, the background sample fit about 31 times, the rotated-frame warp about 5 times, batch photometry about 150 times.

Built around your CPUautomatic

Frame processing is limited by memory speed, not arithmetic, so OriginStack runs single-threaded workers on three quarters of the physical cores, which beat the obvious all-threads layout by 7 to 10%. The release builds run on any 64-bit PC; where a kernel has a faster AVX2 version, it is chosen when the program starts.

Big nights in small memorystreaming

Frames are processed one at a time and the aligned frames kept in memory only when there is room, with a fifth of it always left free; otherwise they go to temporary files read in bands. 500 frames need about the same memory as 50.

Many nights, finished oncemulti-session

Several sessions of one target are stacked separately where the sky rotated too far to pool them, registered onto the deepest, and finished once on the combined stack.

Seven Orion Nebula sessions, 456 frames: 8 minutes 16 seconds, from 15 minutes 12 seconds before this was reworked.

Rerun the finish in seconds--from-stack

Every run saves its linear stack and settings, so trying another stretch or finishing setting reruns only the finish, with the slow early steps cached.

Which one to use