StackingWizard - stack your subs, simply (public beta)

Beta 4.4 - The Workbench - 18 September 2026

• The cloud and haze verdict is judged at cloud scale: smoothed over a hundred pixels, plain sky only, a patch at least 160 px across, a soft edge, so no single cell can draw a square; a walk's thin outer pointings keep no hot pixel or trail. Less by Luke's eye at each step on the Veil and the Elephant's Trunk, not claimed gone: send a stack that still shows it
• The Workbench: a gear beside UI Scale on every page opens seven levers (the cloud verdict, outlier rejection, each frame's sky shape, the canvas's gradient, vignetting from the walk, the twin sample, memory), the wizard's choice first with help and examples under each, kept for the session or forever, one press back, every change named in the receipts
• Siril-converted frames with a stack count of one sort as lights; a finished stack loaded among the subs is called out by its exposure; a panel of one finished frame lays as it is; the mosaic autosave names its own centre
• Play resumes where it paused, the Looking Glass tooltip never outlives its page, the Handoff's long list scrolls, the grades export to CSV, the receipt names the depth, the star finder's neighbour test is tighter

Beta 4.3 - The Thin Strip - 12 September 2026

• Panels that share only a thin strip lay: a ten percent overlap is judged by the stars that can land on placed canvas there, and two placed panels are enough to predict where the third belongs, so a row of three or a 3x2 at ten percent should now lay whole. Proven on a three-panel California Nebula and a synthetic 3x2; if your thin mosaic still leaves panels out, send the receipts
• The levels of a row no longer run white at the ends, and each panel's soft edge is cut before the laying so no dark seam lines
• The grid of faint squares under a hard stretch of a mosaic was the resampling's own grain pattern where a panel is turned against the canvas: each pixel is now sampled twice along the panel's diagonal and averaged, which takes the pattern down by about half for a star 4 percent wider
• The cloud verdict's quiet zone around a bright star grows with the star (Sadr's arc of dark cells)
• The colour pattern door is on every manifest, whatever the header says, and frames stored behind an empty primary header are read whole; StarNet2 no longer crashes on Linux when AstroWizard was opened by the handoff; the receipts describe a mosaic's geometry panel by panel

Beta 4.2 - The Dwarf's Panels - 11 September 2026

• The Dwarf's mosaics stack as mosaics: panels named 'M 31(1)', 'M 31(2)' are read as panels, and a panel that shares one pointing with all the others is laid beside every panel already placed until it finds its neighbour. The Pile's mosaic toggle is always there now - Panels takes named panels, a spread of pointings, or one folder per panel on your word
• The bilinear demosaic is the default again. VNG, new in Beta 4, put a fine maze into the sky of a tracked night shot without dither; VNG and Malvar-He-Cutler stay behind the door
• The screen stretch is the auto-stretch of PixInsight, Siril and ASIFitsViewer, per channel: every channel's sky lands at the same brightness whatever the file, the frame pane shows each frame on its own terms, and the finish of a walk renders as the in-flight view did
• The alignment's memory guard was budgeting three times what a frame needs: a 16 GB machine with a 26 MP camera now aligns four to six frames at a time instead of two or three
• Dwarf owners: stack your mosaic in Panels mode - leave the toggle on Auto when the panel numbers are in the names, set it to Panels when each panel sits in its own folder. The live single canvas you may have seen in the video is the Seestar's walk, which reads the pointing from every frame's header, and the Dwarf's headers don't carry it

Beta 4.1 - The Bright Star - 10 September 2026

• The dark box on the brightest stars of a walk shot at more than one exposure is gone. Two causes: a short frame's gain against a long anchor was capped below what the exposures ask, so its stars sat low while its sky matched; and the canvas's eye for cloud judged the bright star's cell, where frames differ for reasons that are not cloud. The gain now follows the exposure ratio the headers give, and cloud is judged on sky cells only, and only where a minority of a cell's frames disagree
• Everything else is Beta 4, an evening older

Beta 4 - The Live Canvas - 9 September 2026

• The biggest change since the first beta: the mosaic and walk road is rebuilt from the ground up, and every colour stack now takes the new demosaic. A night that stacked fine on 3.2 should stack better here; if one comes out worse, send it - that is what a beta is for.
• One canvas, start to finish: a walk (a Seestar or Dwarf night in framing mode) no longer stacks pointing by pointing and joins the pieces at the end - every frame lands on one growing canvas as the night was shot, placed against the canvas's own stars, and the wizard votes on every pixel at the end at full size. Watch it happen: each frame on the left with its matched stars, its outline in gold where it landed on the right, and a sketch of the canvas saved the moment the frames are all down, before the vote
• The sky levels are solved together, every frame against every neighbour at once, and the canvas keeps the anchor frame's own brightness - the column bands a clear night showed on the way here are gone
• The canvas has its own eye for cloud: where a frame disagrees with everything its neighbours put on the same patch of sky, that patch is cloud for that frame and the vote gives it no weight there. It speaks only where three or more frames saw the patch, and it can never leave a hole
• Named or pointing panels go through the same canvas, with the rig's field distortion measured from the stars the panels share and undone before they meet - Mike's 3x3 lands nine of nine with seams as clean as a single panel's own stars
• Colour frames are demosaiced by the variable-number-of-gradients method (VNG) for the finished pixels: cleaner star cores and less colour fringing, for well under a minute more on a typical Seestar night
• The grading learned from an eye twice: on a walk a frame is judged against its own pointing for what changes across the sky and against the night for the rest, and the trail and softness lines sit a step tighter at every mood - on a blind sample the grading agreed with my own eye on twenty-nine frames of thirty where the old lines let most of the soft and trailed ones through
• Save master after a walk carries the walk's note and the anchor frame's pointing, the same as the autosave

Beta 3.2 - The Gradient Gamble - 7 September 2026

• The grading sees the cloud: every frame is judged as it was shot (calibration is for the stack), and a frame whose star count or brightest-star light falls well under the night's is a suspect - five frames on one real night and fourteen on another that used to pass, every one of them cloud when looked at
• The walk's tiles and colour patches: each pointing's sky is flattened before it lands on the canvas, and the colour gain of every pointing is solved from the stars it shares with its neighbours - proven on synthetic walks with vignetting, drifting gradients and per-night colour, so a real walk is still the test (the gamble in the name)
• Dark flats load as bias: the tile is now Bias / dark flats, and a frame typed or named dark-flat files itself there
• The final vote never dies for want of one band: when memory runs short it votes in smaller bites, same result, with a line in the receipts
• The memory guard now uses the measured figure per frame in flight - fewer frames at once on big sensors with modest RAM, and a finished stack instead of a dead machine
• The pile tiles take a folder again (subfolders included), loose files have their own door, and a frame stamped LIGHT by mistake can be taken anyway; the temperature gap between darks and lights is said out loud; the receipts carry the clock; one star search per frame instead of two
• The Handoff on a mono night: one camera in six filters is no longer called six rigs, panels that never met their mosaic do not travel as a set, and RGB and narrowband stacks never travel together - AstroWizard combines one family at a time

Beta 3.1 - The Walk - 6 September 2026

• The walk: a night whose pointings step across the sky by a fraction of a field (a Seestar or Dwarf in framing mode) is read as one walk - the pointings stack on their own inside, meet on one growing canvas, and the finish shows the canvas alone. One result, one save, one send. Proven on a synthetic walk so far; a real framing-mode night is the test, so send yours in
• The Mosaic toggle on the Pile - Auto, One canvas, Panels or Off, with a line saying what Auto found - so a heavily dithered night is never assembled in panels by accident, and a named-panel night can be told to grow one canvas instead
• A panel that stood aside at placement is tried again once the optics are measured, with the distortion undone (the corner of a 3x3 that came back empty); a mosaic that still lacks a panel says so in the receipts instead of quietly showing a hole
• Sky levels are solved for all panels at once, so the levels close around a ring of panels instead of drifting along a chain - smaller seam steps on the nine-panel test set
• Combining stacks of different exposure lengths: the deepest stack is the grid, and a grid that can host nothing hands over to the next; the star finder no longer mistakes the edge of the data for a line of stars
• Vaonis Vespera frames file as lights; the Pile's notes wrap to the window; long panel lists fold to one line; the cohort tabs scroll past eight; the working-files picker sits on the Looking Glass before you start; "Save the combined stack" once a combine exists; the pile buttons take many files at once; the colour door fits a small Mac's screen

Beta 3 - 4 September 2026 - Magic Mosaics

• The beta runs on every desktop now: Windows, Apple Silicon and Intel Macs, and Linux - the same wizard, the same four steps
• Mosaics have their first light: panels your capture software named (NINA's _PANEL_2 grammar and friends) each stack on their own anchor and meet on one canvas, sky levels matched in the overlap and the seams crossfaded; a night whose pointings sit far apart is asked about rather than guessed at. Panels are placed by consensus on hundreds of shared stars and the optics' distortion is measured and undone; nine 26-megapixel panels assemble in about three and a half minutes. Tell me how yours turns out
• The curved fit: every frame is also registered on hundreds of stars with a second-order curve when that clearly beats the straight fit - frames from the far side of a meridian flip on distorting optics (a RASA, a Hyperstar) used to land a few pixels off at the edges, and now don't. It costs about half again per frame on big sensors; small-sensor nights barely notice
• Multi-night alt-az stacks anchor on the frame in the middle of the night's turn, so the canvas wastes the least edge - measured on the four-night Seestar set that started this (close to an 80 degree turn end to end), the part of the canvas holding at least half the frames grew from 69% to 95%
• Darks per exposure length and gain, swapped per frame like the flats; a ready-made dark library is read by nearest temperature, never blended; and the size guard now trusts the majority of your lights, not the first file it meets
• A colour camera whose files never say so is read from the pixels - the wizard offers both readings side by side, remembers your pick per camera, and "no, it's mono" is a pick too
• Working files stay on an internal drive: the subs' own drive first, never a silent spill onto a USB RAID or a network share, room checked before the night starts, a drive picker if you want it elsewhere, and a slow drive is named before you wait on it. The frames live in 2 GB pieces, so a FAT32 stick's 4 GB ceiling can no longer stop a night, and the cache no longer keeps the night in memory as well as on disk - a big-sensor night stays under a gigabyte
• Hot pixels are found by their stillness - a peak that sits in the same place while the stars move is a defect and never stars again - and the "matched too tightly" warning now names its evidence instead of shouting at a steady night
• Field fixes: finished stacks you already have go straight to the Handoff (a step-1 button, and stacker-signed files recognised), same-colour stacks split only by exposure length offer a one-press combine, Send to AstroWizard never traps you behind a missing safety copy, the frame column sorts into time order, the pixel peeper zooms 1x to 4x, a Dwarf 3's slightly different frame sizes fit, and a night with no recognised lights no longer dies silently on the Pile (the Manifest draws and offers its doors)

Beta 2.4 - 31 August - the mono era

• Mono filter sets arrive in full: the wizard reads every filter dialect your capture software writes, grades each filter, exposure length and night against its own kind, stacks each filter separately to one shared alignment, and flips between the finished stacks with a click
• Flats brew per filter, calibration shot through a sequencer sorts honestly, and finished stacks hiding in a night's folder are recognised by their own naming
• Send to AstroWizard now carries the whole set - it opens Combine Mono Channels pre-filled on the other side
• A night can be fully cleared in one press, autosaves tidy their own older copies, the window opens properly maximised, zoom and pan are smooth at any depth, and the wizard uses many more of your CPU cores

Beta 2.3 - 27 August

• The frame list is rebuilt from the ground up - two thousand frames scroll like twenty, nothing goes invisible, and hovering a name shows it in full
• A UI Scale dial (top left, 90 to 140 percent, remembered) for comfortable eyes on any screen
• Broken headers heal themselves on reading - some cameras write illegal cards that used to stop the night cold
• A frame the wizard cannot read now says “unreadable” instead of wearing a made-up grade
• Dwarf 3 and Dwarf Mini nights sort themselves - lights, nested dark libraries, both cameras’ calibration - and the manifest now names mixed targets, mixed filters and mosaic panels before you stack them
• The finished stack now looks exactly like the live view you watched it build - same brightness, same colours, same smoothness, at any zoom

Beta 2.2 - 26 August

• Every frame stacks exactly once - copies left behind by other programs (Siril and friends) are spotted by their timestamps and set aside, out loud
• The auto-crop is gone - your full picture ships, honest edges and all, and a meridian flip can no longer shrink a night to a sliver. Crop to taste in AstroWizard
• The Pile stops guessing - the four count tiles are now buttons, so you can point each pile at its own folder; frames with no identification are named as unidentified rather than assumed
• The automatic safety copy asks before outliving its purpose - once your real save succeeds, Remove or Keep, remembered
• The finished view keeps the same sky brightness as the live view at any depth, and zooming no longer changes the stretch
• Shift-click ticks a whole run of frames, the mood switch warns before undoing hand-made picks, and the Mac handoff should now work with AstroWizard already open - tell us if yours still stalls

Beta 2.1 - 25 August

• Hot pixels can no longer pose as stars - the alignment used to lock onto sensor defects on soft or defocused frames and stack the night unmoved
• Blank frames and finished stacks stop being crowned as the reference the whole night is built on
• Network drives work: the working files now live on a local disk, and the space needed is shown before you start
• Stacks carry their own details at last - coordinates, focal length, camera - so they plate solve
• Column headings stay put while you scroll, and one wandering frame no longer crops your picture in half
• The wizard explains itself when it sets frames aside, and names a frame that does not belong

Beta 2 - 23 August

• Big nights fixed at the root: aligned frames stream through a disk cache (freed afterwards), memory stays flat at any frame count, the final vote uses every core - measured on the workshop PC: 187 Seestar frames in 2 minutes 31 seconds end to end, where the first beta needed over 8 minutes for the final vote on 60 frames alone
• Every stage shows a measured time left; a Stop button stands while stacking
• Six stars must agree before a frame is placed; frames get a second, harder look before they are given up
• Still heavy: the Looking Glass list past about 500 frames - a lighter list is next

Stacks tonight

• FITS lights from colour (OSC) cameras and smart telescopes
• Darks, flats and bias - brewed into masters and matched honestly
• Several sessions of one target from the same rig, already stacked (by any stacking program, or your smart telescope), combined into one deeper stack
• Meridian flips, field rotation, nights that drift - registration copes

Not yet

• Camera RAW files (CR2, NEF, ARW) - convert to FITS first
• Comets, planets and the moon - StackingWizard is for deep-sky targets only: nebulae, galaxies, clusters
• Mixing stacks from different telescopes (an S30 with an S50, say) - same rig only

Found something odd?

On The Stack page press copy the receipts (in the card at the top) - it copies the wizard’s own log with your build and platform - and email it to wizard@lukomatico.com. A screenshot helps too. Every report is read.

What the grading cannot see

• A tree, a satellite trail or an aeroplane crossing the frame goes unnoticed - we measure whether stars are stretched, not whether something crossed the sky.
• Frames are judged against the best in the pile, so a short exposure next to a long one looks worse than it is.
• “Standard” is too harsh for some skies. If it sets aside more than looks right, try lenient and pick by eye.

The edges, and mosaics

• Stacks ship untrimmed, so a ragged or noisier fringe where frames drifted is normal - it is your data being honest, not a fault. Crop it to taste in AstroWizard.
• Within one pointing, frames beyond the reference frame’s field are still lost; the multi-night anchor now sits in the middle of the night’s turn to lose the least.
• Mosaics expect real overlap between panels (a tenth of a frame or more). A Seestar or Dwarf night in framing mode is read as a walk and grows one canvas, each pointing's sky flattened and its colour matched first; that reading has been proven on synthetic walks and single-pointing real nights so far, so a real walk is the test - send it in, and the Mosaic toggle on the Pile can switch the reading off.

Not built yet

• Camera RAW files - convert to FITS first
• A calibration library: darks and flats kept by camera and temperature, loaded from anywhere
• Pointing at a parent folder and letting the wizard handle each night inside it