What's missing from Siril?

[stark ermine]

One script is already doing that

[mental geyser]

which script is t

[obsidian obsidian]

oh what is it

[subtle night]

This belongs here

[merry perch]

Real though

Its really not missing anything major

Siril + SAS is quite the free combo

[dark sleet]

Why isnt relinearization possible in Siril? 1) stretched stars 2) stretched nebula 3) after not working relinearization

@stark ermine

[stark ermine]

Use the star recombination tool

[mental geyser]

which script does this

[trim steppe]

lock left us hanging 😔

[mental geyser]

litterally edging us

[dark sleet]

Still, it does not work though the pixelmath code is correct(?!)

[stark ermine]

This is native

No script

[mental geyser]

it would be greatly appreciated if you could actually tell us where it is

[stark ermine]

In the star entry of the processing menu

You know, you enter star recombination in the documentation and you'll have all information.

[mental geyser]

The clone stamp tool?

[merry perch]

Just switched to Ubuntu on my main machine and I wonder if Siril is recommended to be installed as a flatpak or an AppImage

in flatpak the dark theme doesn't work

[dark sleet]

There is the script "patch inpainting Tool" which also provides clone stamping

[merry perch]

Ok I got it to work with flatpak override. Still I wonder which one would be best. Linux packages are confusing lol

But now I can finally stack more than 2k mosaic subs in one go and blink debayered frames without lagging thats so awesome

[subtle night]

Would it be reasonable to perform some sort of gradient removal as part of mosaic stitching? I'm still trying to trace the source of the reverse vignetting in this data, but it made me think about solving how this problem may be similar to MSGR.

[stark ermine]

You can perform a background extraction on the whole sequence before

[mental geyser]

how do i get the use drizzle button

ive chaned all settings on that page but only wants to use interpolation

[merry perch]

dont debayer and it will default to drizzle

[mental geyser]

ah

thank you

[wise heron]

i have artefacts arround the star after staking with Siril drizzle 2x ... what could be?

[merry perch]

I installed Siril as a flatpak on Ubuntu; GraXpert (both script and standalone) had a problem that it wouldn't use the nvidia GPU during denoising and other tasks, and after installing some of the dependencies I installed 3.1.0 rc2 so that it would work with the installed dependencies, and the standalone program now uses the GPU correctly, but through the Siril Graxpert-AI script, it still doesn't use it. I heard its because flatpaks dont detect the GPU by default? Is there something I can do?

[stark ermine]

Don't debayer your images

[zenith ridge]

Anyone getting this criptic message on beta4 while trying to run starnet++?

21:49:37: Siril cannot create TIFF file.
21:49:37: Error: unable to save working StarNet input file...

[mental geyser]

do you have storage?

[stark ermine]

Do you have Antivirus? Sometimes they don't want you to save in Tif

[merry perch]

@stark ermine I get " Error initializing application: no display name and no $DISPLAY environment variable" with some py scripts such as AberrationRemover.py or DBXtract, I now use Ubuntu 25.10 with Wayland instead of Windows, Siril is installed as a Flatpak (could have something to do with that?).

you're running on Linux too right?

[split smelt]

I ran on Kubuntu 25.10 with Wayland for a week or so. Siril ran fine but I built mine from source rather than using Flatpak

I've now reinstalled to 24.05 because 25.10 was too new for some stuff I wanted to install

[merry perch]

For a few days I have been wondering what the best way to install Siril actually is for Linux. From what I searched it appears as the virtualization of the flatpak prevents it from using some system resources

[stark ermine]

Flatpak build is not compatible with Wayland (Wayland sucks)

[subtle night]

What are your gripes with Wayland? I know it's becoming the default in more places...

[stark ermine]

Too many things...

[merry perch]

So would you recommend AppImage or straight up building from source?

Also that would explain why hardware acceleration through the GraXpert script doesnt work while it does on the standalone

[stark ermine]

If you know how to compile then go ahead!!

[stark ermine]

Maybe your harward is not supported by onnx

[merry perch]

Should I use the lock042/siril PPA?

Or straight from source?

[stark ermine]

lock042 ppa is the package I maintain

from sources you could have all the update as soon as you pull new sources

[merry perch]

Ok so what changes is how early i receive the updates?

[merry perch]

If I decide to build from source do I have to install the python3 venv package or does Siril just come with its own venv?

I am a beginner so I am just figuring out stuff thanks

[stark ermine]

It will come with it's own

But you need at least to install the python3-venv package

[merry perch]

Ok thanks, also, building from source installs Siril system wide unless prefixed otherwise? Is system wide okay?

[stark ermine]

yes of course

This is how I install it

[merry perch]

great then, built from source makes siril faster than flatpak too?

[stark ermine]

I don't know

but you will have fixes as soon as we publish it

[merry perch]

Ok succesfully installed from source, at first it gave me a " Warning: unable to install or update the Siril python module." but i reset the venv and now it says its up to date. Btw is it normal that by default no .ssf script is preinstalled?

[merry perch]

Ok the display error has been solved but now the ttinker one remain

[merry perch]

Is this still due to Wayland? Any way around it?

[split smelt]

did you install python3-tk with apt?

[merry perch]

No, thats necessary?

[split smelt]

I think I recall doing it to overcome a similar issue

try it?

[merry perch]

Ok I see, I am going to try

[merry perch]

It fixed that error but now I am fighting with this one, i did some troubleshooting and nothing is wrong or missing from my system. For now I saw this happening with DBXtract and AberrationRemover, so probably scripts that want to use ttkthemes

[split smelt]

dunno I never saw that error. Perhaps Google for it?

[merry perch]

I tried, cant really seem to find anything relevant as this specific issue

Could be that I am on 25.10, with Python 1.13 that is very new

[stark ermine]

which script?

[merry perch]

DBXtract

Also AberrationRemover

[stark ermine]

because now, many scripts don't use tkinter

oh this is an issue with your locale

what language do you use?

in your OS?

[merry perch]

Italian

[stark ermine]

because there are some known issue with python and some locales I believe

[merry perch]

So, I should try switching to english?

[stark ermine]

Changing language of Siril won't help

this is something related to the OS

[merry perch]

So, english, in the whole OS?

Chatgpt to troubleshoot has made me go through some loopholes verifying where the C locale came from but there appears to be nothing wrong, so could just be a matter of language?

[stark ermine]

https://gitlab.com/free-astro/siril-scripts/-/issues/37 : some old issues

[merry perch]

Graxpert doesnt give me the issue

But yes it could happen with any script i guess, most work

[trim steppe]

stumbled upon an issue preprocessing some comet stuff, ran starnet on images registered on the comet that had drizzle applied (default settings), opening the sequence in siril that had starnet applied it looks incredibly clipped and stacking them only worsens it, opening the subs by just dragging them in is completely fine though

nvm, its just one of the images thats borked apparently

[broken peak]

one question about scnr, considering that's it clips data and essentialy ruins the image, why is it kept in? does it have some application that makes sense to use it?

[subtle night]

People hate green I guess (their opinion is wrong)

[subtle night]

There's a characteristic bicolor look for SHO that SCNR produces and I'm sure users would really like that to be available

[broken peak]

but wouldn't that also clip the data?

[subtle night]

SCNR? Of course

[broken peak]

mh then it's only for people that don't really care about that

[subtle night]

I'm as confused as you are about SCNR's popularity tbh

[broken peak]

the popularity is easily explained Imo, it removes the green tint at the press of a button and most people are fine with that

I used it a lot when I first started

[subtle night]

Manual color calibration also does that without destroying data

[broken peak]

but it's not at the press of a button

[merry perch]

I guess SCNR comes in handy only when the green tint is extremely stubborn and doesn't go away with CC or curves, so as a last resort

Or when you have color gradients

[subtle night]

On an unrelated note, I feel like TGV denoise would be a good addition to SIRIL. It has some mathematical properties that make it really suited to astrophotography (it assumes the image is piecewise continuous rather than piecewise constant, so staircasing artifacts are eliminated)

[steel grotto]

ive knew scnr destroying data for a bit but i dont see much of a difference in my images so i continue using it instead of white balance

[stark ermine]

Using it on starmask is good

[merry perch]

Does the "Noise" weighting option in Siril help when you have subs for example, from moon nights and some from moonless nights, to stack them in a way that it gives the moon subs the necessary weight to not make them degrade the stack?

is there any extra info on what the main weighting methods do?

[subtle night]

I find weighted FWHM to give me slightly better results; it weights the star FWHM by the number of detected stars, so frames taken under worse skyglow are still going to be weighted lower than those under better skies.

[merry perch]

Good. I notice I get more noise with it though

[subtle night]

I will add that I'm prioritizing sharpness over noise so that may affect my evaluation

[merry perch]

I think I just noticed something weird, not a real issue but something that makes me curious. Previously I had Siril installed as a Flatpak and what I use to notice is that if I opened a sequence and started holding arrow down it would smoothly blink through every frame, without skipping one. Then I uninstalled that and decided to build from source, however I noticed that now this isnt possible anymore. If you hold arrow down on thr keyboard it doesnt really blink subs, it mostly skips through them as they try to load, but they dont, just like what happened for me on windows.

[stark ermine]

that mainly depends of the frame size and the memory used and available

[merry perch]

Guess who got Siril 1.5 instead of 1.4 RC1 by pulling git

Ok yeah building from source without specifying branch guess it does that

[merry perch]

we're already in 1.5 damn

[stark ermine]

You should read the release note

Especially:

Important note for master branch users: Once the stable 1.4.0 version is released, development of Siril 1.5 will begin on the master branch. Users accustomed to using daily builds from master are advised to exercise caution during the early phases of version 1.5 development, as this period may be more unstable than usual. We recommend waiting a few weeks after the start of the 1.5 cycle before resuming daily use of the master branch.

[merry perch]

To which branch should I switch? to do this i delete the _build directory and rebuild it from scratch?

[stark ermine]

1.4

No need to rebuild from scratch

[merry perch]

Not even to remove build to rebuild cleanly? thats not needed to downgrade or change branch?

[stark ermine]

No. Just change the branch and compile again

[merry perch]

great thanks

I notice that every time I compile btw preferences like the gaia catalogue location or installed py scripts reset and I have to set them again, thats normal?

[stark ermine]

No. I compile all the time with no issues

[merry perch]

ok then i'll see if it happens again

I think I get a error opening scripts directory on the terminal

[olive hinge]

amma have to read the releas notes

[merry perch]

Reverted back to 1.4.0 RC1, my Python scripts are succesfully loaded but the default .ssf scripts are absent. But like some other times i get these errors, should I create the directories myself if missing?:
Entering wrapping_idle for function 0x6110574554b0 scripts: Error opening directory “/tmp/.mount_sirilA6RHwe4/usr/share/siril/scripts”: No such file or directory scripts: Error opening directory “/home/mattitech/.siril/scripts”: No such file or directory scripts: Error opening directory “/home/mattitech/siril/scripts”: No such file or directory

[stark ermine]

remove all the lines in the siril prefrences, scripts tab

[merry perch]

Done, now siril doesn't respond

had to kill it, now the same paths show and same error

Edit: the directory of the .ssf scripts for me is in the directory where I built Siril source, I just found out, not where Siril is looking for in the home directory, changed it, now it finds the scripts, just hope it saves that for next compilation

so random but the locale error with the py scripts doesn't happen anymore, very cool

[wise creek]

Following up, has it ever happened again?
If it does, can you pls test with this version of Siril: https://gitlab.com/free-astro/siril/-/jobs/12025374158/artifacts/download
Do not restart the PC and use the memory settings/cpu number as usual. We need to check if the memory allocation (implemented only on some very specific parts) performs as expected with a memory in fragmented state, which is reset when you restart the PC. It should also fix the crash that occurs with drizzle

[trim steppe]

ive had my allocation set real low so not really, could try though

[wise creek]

So to test, you would need to reinstate a memory ratio of 0.9 or smthg, as it was before this bug occurred

[trim steppe]

yeah, ill do it sometime tomorrow most likely

[broken peak]

any idea what causes this error in the continuum subtraction python script?

[stark ermine]

What is the version of SIril you are using?

[broken peak]

1.4.0 beta 3

[stark ermine]

well, update

[broken peak]

mh I didn't know about newer updates, will do

[stark ermine]

beta4 and rc1

[stark ermine]

You must have disabled the automatic update check, because Siril would have told you otherwise.

[broken peak]

I don't think I've ever touched that setting, Idk

updating now, I'll enable the update check if it's not already on

yeah it was disabled for some reason

do scripts need to be updated? the continuum subtraction one is giving me weird results

[merry perch]

Somehow, when turning on show objects option I get this weird visualization. But only after the image has been platesolved on Siril through either the standard solver or astrometry.net, but not when the image is already is solved by astrometric stacking

This is on rc2 and also happened on rc1 and beta 4

[stark ermine]

This is a wcslib bug

I already reported it

You're on debian?

Downgrade wcslib will work

@merry perch

[merry perch]

Ubuntu yes

Ok thanks

[trim steppe]

i forgor 💀

[stark ermine]

Scripts are automatically updated

[safe shale]

After registering frames, the plot only shows up with the converted frames and not with the registered ones
Did I do something wrong?

[safe shale]

Also the stack looks like this

[stark ermine]

known issue fixed in the bext version

[stark ermine]

here you just use a bad Bayer pattern

you should let Siril decide what pattern to choose

[safe shale]

Ty

[prisma valley]

crazy that this has turned into tech support

[trim steppe]

always has been

[subtle night]

I was thinking it might be nicer to have a SIRIL users' thread

[merry perch]

Maybe What's missing from Siril is the friends we made along the way

[olive hinge]

[broken peak]

Yeah I always meant to ask to rename the chat

And always forgotten

[subtle night]

I mean I'm personally interested in SIRIL development so this thread (if it stays on topic) is useful to me at least

[broken peak]

it could be just named "siril chat" or something like that

[subtle night]

Any recommendations for building a patchwork mosaic in SIRIL? (Stitching together a bunch of images taken by other people - different palettes and processing styles)

[subtle night]

Also @stark ermine how feasible would it be to interface a different programming language with SIRIL? I'd really like to work with some tools I'm writing in Julia in combination with SIRIL

Obviously there's the Python tooling but I'm also the author/maintainer of several Julia packages :)

[olive hinge]

make the python pkg launch the julia one

[subtle night]

I mean that is probably workable, but I'd prefer something less roundabout

[stark ermine]

Hello. I'm not sure we want to use something else. Siril is coded in C/C++ and we use python for scripts

THis is already very complicated to maintain something consistent on all platforms

I don't think that using another language would be easy and helpful

[subtle night]

Yeah, I figured there's probably a large amount of developer effort involved

[stark ermine]

yes 🙂

[subtle night]

I'm working on some denoising code which I would love to have available in SIRIL (TGV denoising that assumes Poisson statistics). Tbh I would love to contribute to the SIRIL codebase but my skills in all of these languages are kinda limited (I was taught C from a chemist who did not know how to malloc, free, or implement a multidimensional array )

[olive hinge]

[subtle night]

I have flashbacks of the code we used for our research. It's not pretty

[olive hinge]

is that a block of variables that have had their names truncated to be impossible to read without knowing what they do

[subtle night]

I am actually trying to implement the algorithm described in this issue in a Julia package, so perhaps this may be useful to Adrian

[subtle night]

There are 9 separate variables for matrix components instead of just using 3D matrices available in GSL...

wait until you spot the 45 unironic goto statements

[olive hinge]

[olive hinge]

which line is that

[subtle night]

Just use Ctrl+F on "goto" and you'll find them

[olive hinge]

yea

ive found tham

also im on mobile rn

i have seen something about double electrons or something and a periodic table

what is this about?

dft chemical pressure analysis

[trim steppe]

tried with one dataset and it was fine, both current and maximum modes on drizzle, stacking with maximum was fine too, was using a ratio of 0.95 for memory

was a rather small dataset though, will try with the version of siril i had installed before as well

[trim steppe]

it didnt crash in the normal version

[stark ermine]

Working on a new script

This is about HR diagram

Could someone with PixInsight use the script on one of their images (ideally a globular or open cluster), send me the image in a private message, and show me the result in PixInsight? That way I can compare

[subtle night]

Would you find it useful/interesting to have an IRG cluster image? (I don't have PixInsight, but I do have some HR diagram data that my friend got for the RGB version)

https://media.discordapp.net/attachments/1310324748281974854/1430251953761288252/HRDiagram_DoubleCluster.jpg?ex=69214f2e&is=691ffdae&hm=f337f3eb7b1b4b876593e7381079256876292bb9cca6252e414f3d827df57721&=&format=webp&width=1558&height=1336

[stark ermine]

Yes please

I would like to compare both results

[subtle night]

Files are getting uploaded right now

[stark ermine]

🥁

[subtle night]

Still uploading but:
Double Cluster (IRG, RGB, and G2V-calibrated RGB)
Also available
M13 (IRG, taken from a light polluted area so there are gradients)
M24 (IRG, also includes the H-II region Sharpless 41 throughout)

[subtle night]

Double Cluster has finished uploading

[stark ermine]

Which file was used to create this?

[subtle night]

Uncalibrated RGB version

[stark ermine]

ok. What are the filter and sensor ? (needed for SPCC 🙂 )

[subtle night]

IMX533 mono & Touptek RGB

[stark ermine]

thx

[subtle night]

IR data was taken with an SVBONY 685 nm IR pass filter and it reuses the Touptek RG data

[stark ermine]

this is the result

With apparent magnitude

I think we're good 🙂

[subtle night]

I didn't even realize how many white dwarfs there could be

The Double Cluster data is CC0, I'm releasing it to the public domain

[stark ermine]

thx for your share

So, in the coming days, I think Siril will have the HR diagram script too

[subtle night]

Actually do you mind if I share the diagram on my social media?

[stark ermine]

Please do, but the UI dialog all around may change a bit ^^

Wait I can share you only the plot if you want

[trim steppe]

when we getting something like mars in siril

[subtle night]

Thanks!

[stark ermine]

Is this thing really worth the effort? I've heard that the results aren't great.

[stark ermine]

If I use the same axis scale:

[olive hinge]

where the flippy this come from?

[stark ermine]

??

[olive hinge]

what is it

and what software exists it

[stark ermine]

hertzsprung-russell diagram done with Siril

[olive hinge]

i didnt know siril could do that xD

[stark ermine]

I'm writing a script allowing that

[olive hinge]

nice!

[stark ermine]

Good news, the script is now available

[olive hinge]

amma get

[prisma valley]

HR diagram my fav diagram

[merry perch]

Amazing. Sorry if its a stupid question but does this work simply by analyzing the stars' color and flux based on the image data or does it need an astrometric solution?

[stark ermine]

astrometric solution is also needed

it compares stars between gaia catalog too, in order to know the star distance which is important for absolute magnitude

[floral plaza]

where can i find this tool? :o

[stark ermine]

like all other scripts

[floral plaza]

I figured.. perhaps a hint on under what tab/folder it might be?

[stark ermine]

??

Scripts->get Scripts

it is easy to find

[floral plaza]

alr ty! assumed it came with the latest siril version

[desert tangle]

Im getting this error

21:31:32: Starte Script /home/lukas/siril/scripts/HertzsprungRussell.py
21:31:35: Error: failed to retrieve FITS keywords: Received keyword data size 1465 doesn't match expected size 1235
21:31:40: Error: failed to retrieve pixel data: Timeout while receiving data
21:31:40: Fatal error: 'NoneType' object has no attribute 'ndim'
21:31:40: Traceback (most recent call last):
21:31:40: File "/home/lukas/siril/scripts/HertzsprungRussell.py", line 1218, in main
21:31:40: window = HRDiagramDialog(siril)
21:31:40: File "/home/lukas/siril/scripts/HertzsprungRussell.py", line 823, in init
21:31:40: if fit.data.ndim != 3:
21:31:40: ^^^^^^^^^^^^^
21:31:40: AttributeError: 'NoneType' object has no attribute 'ndim'
21:31:40: Python Prozess (PID: 29330) mit Status 1 beendet

[stark ermine]

what is your version of Siril ?

[desert tangle]

1.4.0-rc2

I just updated

[stark ermine]

flatpak?

[desert tangle]

no, fedora

[stark ermine]

reset the venv in siril preference please

[desert tangle]

where do I do that?

[stark ermine]

in the Siril preferences, Scripts tab

[desert tangle]

It worked, thank you for your help!

[merry perch]

@stark ermine Is it still helpful to translate strings on weblate now near official release or am I late?

[stark ermine]

sure it is

it is always useful 🙂

[stark ermine]

New version available 🙂

[subtle night]

Which cluster is this?

[stark ermine]

Probably not the best image for scientific study, but I take what I have ^^

[safe shale]

It only works with linear data, right?

[stark ermine]

yes

and broadband

[olive hinge]

do it require ir cut?

[subtle night]

Are you shooting mono with RGB filters? I think that should be sufficient

[olive hinge]

im osc

i dont even have a tracking mount yet, why would i invest in mono

[subtle night]

Ohhh well I think you should be okay if you can use SPCC on your data

[olive hinge]

how you spcc data?

what even spcc

[subtle night]

Spectrophotometric color calibration, it's in the SIRIL docs

[olive hinge]

oh

dont you require a filter for that or somethings

or am i just going crazy

[subtle night]

Not necessarily

[olive hinge]

ah

okie

[olive hinge]

what if im doing rgbir?

do i just deal with not having spcc

[subtle night]

I am honestly not sure in that case since I only shoot mono

[olive hinge]

what if im doing irir

is that same?

bc ill be stacking ir data as mono images

[subtle night]

You may need to use some custom filter data depending on what's available in SIRIL, I'm not sure

[olive hinge]

oka

[subtle night]

https://media.discordapp.net/attachments/1443022857691988098/1443022859609051146/orion-IRG-2panel.png?ex=69278f43&is=69263dc3&hm=6222c28f3dbfb1e9cb538ba5fce90c0892b034fc0a1a10ab949e383429a16ba3&=&format=webp&quality=lossless&width=3192&height=1712

I would love to SPCC my IRG data though :)

[olive hinge]

what colour is it?

[subtle night]

green due to H-alpha being in the green image channel

[olive hinge]

a

thats cool :D

[subtle night]

thanks to @chilly root for pointing this out

[broken peak]

Is that ai generated?

[subtle night]

The tutorial seems to be done by a real person

Who does not know much about how to use SIRIL

[broken peak]

Cool

[subtle night]

So it's still slop

[broken peak]

The thumbnail is absurd, that guy looked like Sainz at first

[subtle night]

...you're goddamn right

[subtle night]

@stark ermine under some circumstances it seems like the saturation stretch functionality in GHS can alter the hues of my image
Original (left) vs altered (right)

I mean it looks cool but that is not the intended effect lol

[floral plaza]

"Hey siri, start NINA and begin livestack B)"

[broken peak]

Hey siril, clip the blacks

[hard cypress]

🎼 🎶🎵 Just rush and push it all, clipping all the blacks 🎶

[merry perch]

https://tenor.com/view/jarvis-erase-gif-1159901542694943356

[merry perch]

Should RGB equalization in the stacking tab be checked if I plan on color calibrating the image?

[stark ermine]

does not matter

you will apply a background extraction before color calibration

the bge apply a RGB equalization too

[merry perch]

So whats it there for?

[stark ermine]

in order to make users happy with a image that is not like Shrek after stacking

[merry perch]

lol

So it assumes you wont process the image?

[stark ermine]

no, not necessary. I always check the option

but if I forget, this is not a big deal

[subtle night]

"What are you doing in me swamp?!"

[olive hinge]

downloading spcc catalogue i think

the window crashed when i clicked a button

so i guessed which place the download button was in

and something is downloading

[stark ermine]

hum. I don't understand what you are saying

you mean, gaia local catalogue?

how did you download it?

[olive hinge]

i pressed the button to the right of the button that i pressed to download (that dissapeared bc interface crashed)

[stark ermine]

ok, you are using remote catalogue

[olive hinge]

which caused it to die

[stark ermine]

while Gaia is under maintenance now.

you should use local catalogue

we have built a catalogue for this case

[olive hinge]

thats what im downloading ._.

so, i open the offline dowloader

[stark ermine]

oh ok

[olive hinge]

input everything for it

[stark ermine]

did you use the script?

hum

[olive hinge]

then to the right of the download button theres like a visualise button or something

that crashed the ui

so i guessed where the button to download was

but it worked somehow

[olive hinge]

yea!

that thing

it didnt open

[stark ermine]

ok. No crash for me

[olive hinge]

buh

im just special

[subtle night]

Not a SIRIL feature, but I would love to see a gallery of user images on the siril.org website

[stark ermine]

Could be an idea

[merry perch]

Yeah its not like only Pixinsight is allowed to flex

I see to many "Siril vs Pixinsight" comparisons where the Siril image was processed in 30 seconds and with zero familiariaty with the program and then a meticulous Pixinsight version

Its so bad many people think you have to get Pix for "nice colors"

[olive hinge]

my dog a siril supoorter btw
|
V

[stark ermine]

processed with Siril

of course you can have nice colors with Siril

[merry perch]

great example

[olive hinge]

it looks like a garlic

[stark ermine]

We will work on it before the end of the year I think

[trim steppe]

when we getting an official siril discord server and not just a thread in ab

[subtle night]

I'd join

[subtle night]

Though of course there are user forums: https://discuss.pixls.us/c/software/siril/34

Though I will say dark mode should be the default for the forum. It's not fun to get flashbanged when you're troubleshooting something in the dark

[stark ermine]

Once you set it as dark, it will stay dark.

[stark ermine]

New python script for an automatic stretch better than the statistical one 🙂

This is the easiest way to have nice stretch and nice preserved colour

[subtle night]

Can you apply this autostretch to the image?

[stark ermine]

This is the purpose to the script

Apply an autostretch to the image

and a better autostretch than others

[subtle night]

Oh sorry I'm asking the wrong question

Because I thought this was using GHS as part of the preview functionality

Either way, I would like to consistently apply stretches to my images so I will give this a try

[split smelt]

For instant results, pick "read-to-use" mode and don't turn Log D up too high or colours will be too saturated

[subtle night]

Did you develop this script?

[stark ermine]

I press the button to compute it automatically :).

[split smelt]

Not me

[split smelt]

I tried that but it computed too high. Had to turn it down a little

[stark ermine]

oh ok, good to know

maybe you can reach the dev to tell that

[olive hinge]

siril

[olive hinge]

woaa

[split smelt]

I'd like to test it more before I feed back. Perhaps it was just an oddity with my data

this is what I ended up with, with very little effort

[stark ermine]

Very nice picture 🙂

[subtle night]

CALL THE FIRE DEPARTMENT

SIRIL 1.4.0 JUST DROPPED

Thanks @stark ermine and everyone on the Siril team for your work and listening to us all yap about what we want regardless of how contradictory it may be

[olive hinge]

car moon

@stark ermine how do i use siril for lucky processing?

bc i do lucky imagery

and i wanna be able to do rejection stacking well

and use a video file instead of indeviduals

[stark ermine]

Use SER

All other video formats are bad

And with SER you use Siril in the same way you use it with standard FITS

This is the power of sequence

[olive hinge]

woa

is it possible to add rejection stacking to a script?

because ive made 1.2x drizzle scripts and stuff

and bayer drizzle without dbf

but i haven figured out how to rejection stack

[olive hinge]

https://youtu.be/ZIw4TN1XAvc

woww

qlready bido

[stark ermine]

Of course it is. Usual scripts have already rejection.

[olive hinge]

huh

[trim steppe]

[olive hinge]

hey, stop screaming :(

[olive hinge]

is it updated on S.u.S.E repo?

[stark ermine]

Siril? I don't know

I don't manage this repo

[olive hinge]

ah oker

ill find out after it finish dup

[olive hinge]

trusnout i had it through flatpak

mai bad

[olive hinge]

buh

tried to use vera lux

[olive hinge]

had to reset venv

it worky now

[olive hinge]

i didnt get brown with my haii data

[stark ermine]

haii ?

[floral plaza]

https://tenor.com/view/wave-emoji-smile-wave-smiling-waving-emoji-haii-gif-7959794316942733877

[olive hinge]

hydrogen alpha + ir ir

its colour correctly to hwo it mean to be is what i mean

[olive hinge]

irgbir 🤤

@subtle night

i got the 685nm out

[broken peak]

I think you need to make a tutorial on your process, I can never get anything really good but your images on telescopius are great

[stark ermine]

I wrote a book, currently translating it in English

[olive hinge]

do you like my igbi onion image

[stark ermine]

colors are ... funny

[broken peak]

Oh nice, can't wait to read it then

[stark ermine]

https://siril.fr for the French version

[broken peak]

Any eta on the English version?

[stark ermine]

Q1 2026

I hope

[olive hinge]

infrared momen

[olive hinge]

@stark ermine can siril take 8 bit video ;-;

i didnt realise it defaulted to 8 bit and not 16

[stark ermine]

yes of course

SER file?

[olive hinge]

yea

fircapture has dementia and keeps forgetting some settings 😭

[merry perch]

@stark ermine Why do I frequently get weird corners and vignetting when using linear fit clipping (default settings)?

Doesnt happen on mosaics though

[zenith ridge]

That is from stacking. Probably there you only have a few subs toward the beginning or end of the session

[merry perch]

It only happens with linear fit, if I use Winsorized sigma clipping it doesn't

[stark ermine]

these methods are quite differents

I never seen that however

probably due to a gradient in each sub

[hollow garnet]

@stark ermine The new mosaic tool is incredible! it stitched Hα perfecly in a matter of seconds, though it's having problems with OIII and SII (mainly SII as you can see). I think the issues are caused at least partly by SNR differences when averaging in the overlap, instead of just overlaying them. Any ideas on how to proceed?

(viewing in pix because of the 32k pixel limit in siril)

[stark ermine]

Sure, you could apply a background extraction on each panel before

[hollow garnet]

I already did

maybe not good enough ¯_(ツ)_/¯

but OIII is done with MGC so it really should be good

[stark ermine]

how many pixel for the overlap ?

did you check the normalisation overlap?

[hollow garnet]

i did

[hollow garnet]

about 1k

[stark ermine]

try the max (2k)

[hollow garnet]

wait then i'm not sure i understand. I have 1k pixels of overlap between the frames?

oh wait feathering

[stark ermine]

oh, I was speaking about border feathering

[hollow garnet]

i only did 56 lol... lemme try 500

not sure it if it helped that much

(starx takes 10 minutes to run so not starless this time)

SII btw

[stark ermine]

try the max

looks better for me

[hollow garnet]

i think it's the stars, but i'll try

[hollow garnet]

:/

[subtle night]

@stark ermine on a related note, I find the denoiser has some trouble with panel overlap (it smooths the overlaps excessively). I figure this could be mitigated when masking arrives in 1.6 (IIRC) but in the meantime do you have any tips to reduce the effect?

(This is worse when you have less integration, if the panels already have sufficient integration on their own it's less of a problem)

[stark ermine]

For your information, I've finished the translation of my book about Siril (https://siril.fr). It should be available in Q1 2026 in USA (even if customs duties are a pain in the ass) and of course in other countries. I hope you'll enjoy it. French users have responded very positively to the book's release. I hope it'll be the same for you.

[gusty spoke]

Is it a guide and explanation of tools in Siril?

[stark ermine]

yes

[gusty spoke]

cool il buy it once available in the US

[modest heron]

hi everyone, ive been trying to install the SPCC catalog lately, but so far i keep getting an "installation error: failed to download file""

does anyone know what could be the cause?

[olive hinge]

also getting that

idk the fix tho

[stark ermine]

https://zenodo.org/records/14738271 : you can manually download the files

[modest heron]

i guess that by doing it manually, i have to download the entire catalog

[obsidian obsidian]

I didnt know siril was made in my country

[stark ermine]

You're French?

[obsidian obsidian]

Oui !

[mental geyser]

oh non

[zenith ridge]

1. There is a new version of statistical stretch with more parameters exposed

2. I miss some option to do HDR on the highlights (SAS pro has some tools that are nice) and also some tool to add more contrast to the shadows.

3. Blumish tools (for comet stacking, to remove the residues from stars)

4. An option to cleanup all the p_, r_ files, symlinks, etc after the final stack image is generated. Cleaning manually is a bit tedious and leaves lots of crap in disk.

5. If it is about asking, possibility to have previews (to test time demanding processes like AI stuff on smaller regions) and even multiple files opened and processed/linked simultaneously.

6. Fix the random crashes when doing RGB combination. When you open a new image to assign to a channel, sometimes the entire app crashes en macosx.

Maybe some of these are impossible with the current architecture, and maybe some of them are there but I just missed.

[stark ermine]

There is a new version of statistical stretch with more parameters exposed

OK, but Veralux stretch is better anyway

An option to cleanup all the p, r files, symlinks, etc after the final stack image is generated. Cleaning manually is a bit tedious and leaves lots of crap in disk.
https://siril.org/faq/#could-you-add-commands-to-delete-files-or-sequences

If it is about asking, possibility to have previews (to test time demanding processes like AI stuff on smaller regions) and even multiple files opened and processed/linked simultaneously.
-> Did you try ROI feature? and for the other point the multiwindow script

Fix the random crashes when doing RGB combination. When you open a new image to assign to a channel, sometimes the entire app crashes en macosx.
Fixed in the 1.4.1 coming next

I miss some option to do HDR on the highlights (SAS pro has some tools that are nice) and also some tool to add more contrast to the shadows.
Siril has a HDR multiscale script

[stark ermine]

There is a new version of statistical stretch with more parameters exposed

Updated. But I really think that VeraLux scripts are far better

[olive hinge]

[zenith ridge]

Well in my experience veralux is not far better than statistical stretch. It just saturates too much the images and exacerbate any mis-color calibration possibly present in the image. I keep telling myself to give it a chance, and every single time I come back to statistical stretch because for me it simple works much better

Thanks for the tips and updates !

[stark ermine]

You really need to test the last VeraLux version

And, no, it does not saturate colors while MAS does it (with a slider)

VeraLux preserve color vector

if SPCC is well executed right before

[stark ermine]

Veralux stretch has been released in its 1.5.0 version

[merry perch]

@stark ermine sorry for the ping. I can't remember exactly what I did to build siril from source on Ubuntu, but now I realized for all this time Siril has been installed in /usr/bin and not /usr/local/bin. Is this a bad thing for the system or is it supposed to be installed this way?

I need a little clarification on what you mean that you install it system wide

[stark ermine]

This is not a big deal for me

No problem

[merry perch]

Okay. I was simply reading its bad practice to put your binaries with system ones, because I know that /usr/bin is where apt works. I have the _build directory still so I just uninstalled, wiped and reinstalled again using the /usr/local/bin prefix

Worked fine so this is where its at I think

[stark ermine]

new AberrationRemover version in Siril.

• New UI
• New model

[merry perch]

https://tenor.com/view/this-peak-is-this-peak-so-peak-peak-it's-just-so-peak-gif-1360442599844596686

Its so cool that with AR and CosmicClarity i dont need to beg others for blurx as much

[olive hinge]

whaaaahhhhh???

[stark ermine]

Yep. This is very good :). You cans spread information to other channels 😉

[olive hinge]

gah

im using the old oneeee

i may have to reprocess 😭

[trim steppe]

wonder if this one will leave those tiles in images too at times

[stark ermine]

nope it has been fixed

[olive hinge]

it also insnt adding stars in anymore

it just takes them away

(star had a 1 pixel seperation and it eated it)

[stark ermine]

?

do you have screenshots ?

[olive hinge]

this the before

then it cronch

went omnom

its no big deal tho tbh

makes sense why it ate it

[stark ermine]

ok. This is with the new model?

[olive hinge]

yus

[merry perch]

I noticed a bit of a similar issue. I tested it on two images for now, on a specific horsehead image it does this, also while messing with some structure in the flame nebula, it happened in this image, I tested it in IC1805 and it was much better though. I am keeping the old model availabme just in case, and I reported it on github.

This has a lot of potential

I had to reduce strenght to solve it, and the author clearly said that if you do that you could count that as an issue. Thats why I reported

[zenith ridge]

when I try to do the registration with distorsion modeling, I get super weird results:

[stark ermine]

Here this is the result of stacking with rejection

[zenith ridge]

without that distorsion it works better, slight artifacts near stars, but better

[stark ermine]

check your registered sequence first

I always use distortion correction with no issue

[zenith ridge]

not so sure. It looks more like it is not correctly aligning the stars

[stark ermine]

so probably there is something wring in your data

[stark ermine]

yes, and rejection do the artifacts

[zenith ridge]

without distorsion correction it works fine, my data is ok. A bit of tilt/coma on the right side, but not dramatic.

I always keep this setting disabled because it never worked with any of my rigs. But now I wanted to do some mosaic and thought I rather get it sorted

looks like changing the windsored sigma clipping to another algorithm and plate solving manually the reference image helped

[stark ermine]

I can't say without data

[olive hinge]

ive found that i can mosaic just fine without it

[stark ermine]

That depends of your distortion

[olive hinge]

[subtle night]

Have you checked the plate solve solution/distortion map? I had a similar result once and it was due to a lack of stars, but clearly this isn't the problem here

[subtle night]

This is why I'm a glorious Ritchey–Chrétien enjoyer. It's designed for minimal distortion

[zenith ridge]

I think the key was to select an adecuate reference image and manually plate solve it.

[stark ermine]

what you mean by, manually plate solve it?

[zenith ridge]

Select the image in the frame list, click on astrometry and trigger a plate solve. I thought it was doing it automatically, but this seems to make a difference

[stark ermine]

hum. To register with distortion you need to choose an image and apply astrometry yes

https://siril.org/tutorials/mosaics/

[stark ermine]

Good news: https://siril.org/2026/01/a-brand-new-star-removal-script-comes-to-siril-zenith/

[olive hinge]

betta than starne

?

looks it

[stark ermine]

Test it

[trim steppe]

well uuh

chomped quite a few galaxies and the core of cigar

and just left one strip undone 😭

[trim steppe]

not cookin

[mental geyser]

[dark sleet]

It is also removing galaxies bruh

[merry perch]

Honestly though its about time something other than starnet comes in

I still need to test it

[trim steppe]

still the first version ig, let em cook

[trim steppe]

waiting for starnet v3 🙏

[olive hinge]

i think you missed a bit on the side

[trim steppe]

ion know why that happened 😭

[merry perch]

Its coming?

[trim steppe]

id hope so, i remember seeing some talk of it some time ago

[stark ermine]

yes we try to understand this issue. Because I can't reproduce

but you should see that background is far better than starnet

[stark ermine]

Here are the known issues we are investigating about the new star removal script:

https://gitlab.com/free-astro/siril-scripts/-/issues/91

If you experienced other issues, please let us know.

[dark sleet]

When I ran it on M31 it removed M32 and a big chunk of M31s core

[stark ermine]

WHat if you change Tile?

Well, we will ask for images soon. It will help to improve the beta

[olive hinge]

om nom :3

[dark sleet]

Before and after

In the starmask you can see it even removed some of the structures in M31 (especially the core)

[stark ermine]

Yes, but check with starnet. It does the same. All starless algorithm remove a bit of signal, especially in galaxy core

[merry perch]

Performs slightly better than starnet on large stars imo

I still didnt have time to test it myself

[olive hinge]

hey, you missed a bit on the side

[dark sleet]

Thank you for pointing that out. I totally missed that

[olive hinge]

[subtle night]

Is there a way to set the console font? It's showing up as proportional for me rather than monospace

(I switched from Windows 11 to Arch Linux/KDE)

[stark ermine]

it should be forced to monospace

[olive hinge]

YOU USE ARCH NIW??

also, eww arch

[subtle night]

I actually have been using it in other places but I switched my desktop over to it

[olive hinge]

isee

what made u finaly switch?

[subtle night]

Not here at least

[subtle night]

Microslop and its AI crap

[olive hinge]

isee

[subtle night]

Broken updates and all

[olive hinge]

ohh yeaa they now release constant broken updates dont they!

like once a week or something right?

[subtle night]

Worse: Windows kept attempting to reinstall 25H2 every time I booted my PC. While it was already on version 25H2

I found the fix: apparently I had to re-enable telemetry (!!!)

Apparently it had something to do with the Windows Insider program, but I never enrolled my desktop in it

[subtle night]

I'm surprised you didn't flame me for using KDE tbh
What are you on, Gentoo?

[olive hinge]

im on opensuse, i did used to use my own distro i made from scratch

used it for like 3 years

before deciding it wasnt worth it anymore

[olive hinge]

😭

wankers the lot of them

[olive hinge]

also, not flaming you for kde bc i know you used windows

just wait 3-4 months then i will

dont worry

[subtle night]

I wanted to try hyprland but I heard it was not supported on Nvidia GPUs

If I was using GNOME you should flame me though

Already starting to customize SDDM

[olive hinge]

hyperland dev is a twat

hella transphobic and stuff

[subtle night]

Ohhhh I didn't know that...good reason to avoid it then lol

I know some people out there are gonna say "it's just code! who cares?" but it does matter when you're trying to actually work with the people who write your code

[olive hinge]

yea

[subtle night]

What WM/DE are you using?

[olive hinge]

usually i3wm

[subtle night]

Not using Wayland?

[olive hinge]

but im on kde currently

im test running

[subtle night]

I've just stuck to KDE because it's consistent, customizable, and Valve's really brought it into a renaissance

[olive hinge]

im not used to it no

so ive only recently switched

within last 3 months

[subtle night]

Are you using i3 with Plasma or just KWin?

[olive hinge]

just kwin

i3 w plasama is pointless

[merry perch]

As far as Drizzle goes, what droplet model would you use between Square and Gaussian and why (considering potentially thousands of short seestar exposures)? I wasn't able to fully grasp the pros and cons of both from docs

[subtle night]

Square is flux-preserving (important for photometric applications), but Gaussian might give you better star shapes. I haven't had the chance to do much testing, but the difference has been relatively small (at least when it comes to how good the stars look).

You can also use the Turbo kernel for speed if you're operating in EQ mode. It assumes that the rotation between frames is negligible, so it probably won't be a good idea if field rotation is significant.

[merry perch]

So I assume square works better for SPCC in theory?

[subtle night]

Eh, I'm not sure if it makes that big of a difference there. The flux preservation is more so about getting accurate magnitude measurements, so it's more important if you're monitoring star variability

For SPCC it's more so about the ratio of luminosity across bands

[merry perch]

Ok thanks for clarifying

[olive hinge]

suqare dosen effect star shapes imo

[merry perch]

Ok I tried Syqon starless

Very good

Compared to starnet I got mixed results. Like for example Starnet did a better job of removing the huge Alnitak, while Syqon left it more smudged. On the rest they are either on par, or in a specific instance Syqon gave me a cleaner image while starnet would leave a halo or two

Very promising

Also I like that its GPU accelerated, could be my default

[trim steppe]

yeah the gpu accel is nice

[stark ermine]

Calling all astrophotographers!

We’re improving Starless’ Zenith model and need your linear stacked images (16-bit or 32-bit, FITS or TIFF). Even low-quality files are welcome—real-world data helps us train better! Your files will be deleted immediately after processing, and every contribution makes a difference.

Upload your files here:

https://www.dropbox.com/request/IjgrGb7za64dG0r6Zetn
Thanks for supporting the community! Let’s push the limits of astrophotography together.

Don't be shy!!!

[olive hinge]

what if i upload all 500gb of stuff i have

[olive hinge]

ive sent 3 fits

moderately bad data too so it could help

(only one image is good)

[subtle night]

Is there any way to specify that they are welcome to retain my data?

[split smelt]

Starnet vs Syqon on a mono Cone image. Same process for both - star removal, Graxpert denoising

[dark sleet]

Which is which?

SyQon on the right?

[split smelt]

yeh, Starnet left traces of the original stars

[dark sleet]

It removed a part of Hubbles Variable cloud tho

[split smelt]

Is that the bright bit at the bottom?

[dark sleet]

jup

[merry perch]

About to send the most unholy Seestar elongated star stacks known to men, the model will have fun

Stuff like this for the record

[dark sleet]

What do you need my name and email for

[olive hinge]

just write

mrooowwww

as name

and use tempmaol

[dark sleet]

Nvm. I trust you. You cant do that much with only my Mail Information

[subtle night]

I think it's a good question. It would be nice to be able to submit anonymously

[merry perch]

Could be a Dropbox thing

I just got the email back saying i sent the files

[split smelt]

just in case you forget that you sent them 😛

[dark sleet]

I dont get how to stack my Images without Scripts. I always get a shitty result

[stark ermine]

Siril has plenty of tutorials and even a documentation :).

Soon, I will even publish a book in English 🙂

[dark sleet]

Yes, I know. But all the tutorials are for older Siril versions and they somehow just don’t work. I always get a very very dark image in which all the background had been set to 0 (it’s black in histogram). Only a few stars a visible

[stark ermine]

Absolutely not, or you are watching unofficial tutorials

siril.org contains updated tutorials, but yes, they are not videos

[dark sleet]

I will try them thx

But in the tutorials it is all done via scripts. I want to do it manually

[stark ermine]

No, there is at least one tutorial explaining manually

https://siril.org/tutorials/tuto-manual/

[dark sleet]

what is this?

the image after calibrating

[stark ermine]

without all the logs (ideally in English)... difficult to say

[mental geyser]

gonna need the english translation

[dark sleet]

It’s the one I marked

I just see that Siri told me during calibration with darks that the images contain 99% negative pixels. But I have a dark library. That makes no sense

[mental geyser]

if youve 99% negative pixels that means your dark frames havent been applied properly

[dark sleet]

But I haven’t done anything wrong

And even the scripts are messing it up

[mental geyser]

that means theres an issue with your darks

[olive hinge]

your darks are fucke

[olive hinge]

its never user error

[dark sleet]

Ah bruh

Gain 100 Offset 316

[stark ermine]

So you have the issue. Your dark is not good for your light

calibration is pretty simple. This is just subtraction

If your dark has a higher level than your lights... well you will have pleinty of negative pixels

[olive hinge]

gwt outtt

[dark sleet]

My lights are with Gain 300 and Offset 50. My other darks have the same offset but Gain 100. Why did I got 99% black pixels too then? After stacking the image looked fine tho

[subtle night]

Did you apply biases and darks at the same time?

[dark sleet]

I tried with and without

[split smelt]

why are the gains different?

[dark sleet]

idk. Apparently I just gave Nina the wrong numbers

[dark sleet]

Worked now

[split smelt]

What did you do differently?

[dark sleet]

new darks

[split smelt]

great 🙂

[subtle night]

In the future it might be worth trying to just subtract a bias/bias level for troubleshooting

[merry perch]

I seem to have noticed drizzle intermediate registration files have gotten heavier since 1.4? or is it just me? high res drizzle files are disabled

this is 2x drizzle for instance, the original file is 1080*1920, when registered it is turned into 24mb at 1x drizzle

just wanted to know if the specific file sizes are normal, I know drizzle takes more storage

the raw CFA frames are 4.2mb

[olive hinge]

that makes sense uea

sorta

wait

4.2*3 = 12.6

idk

makes some sorta sense

[merry perch]

2x drizzle is 4x the file size afaik

still wouldnt get to 99mb, thats why am asking

[mental geyser]

you've doubled the image resolution so it would be double the file size

[olive hinge]

4* the pickles tho

[merry perch]

yeah but the image in total has 4x more pixels, like you doubled on each dimension

[olive hinge]

then bc its debayers its 3 channels

so triple

it shoud be like 50mb

weurd

[merry perch]

yeah how does it get to 99

1.5x drizzle woth 10k frames nearly totaled my drive

seemed abnormal

[merry perch]

could it be that the cfa frames are 16 bits and then siril turns them to 32 bits

[olive hinge]

yea

[olive hinge]

ohhh

that will do it!

[merry perch]

alright mistery solved ig, but why do i remember it being less. or maybe Im tripping

[olive hinge]

that brings the file size to 100mb

so that makes complet sens

[merry perch]

I never stacked 10 thousand frames all at once with drizzle

[olive hinge]

i think it didnt use to auto make 32 bit

[mental geyser]

yea i believe by default siril does 32bit files

[olive hinge]

[merry perch]

with 1.5x, registration took 570GB

left me with 64 crispy gigabytes on my 1TB drive

[olive hinge]

hey that used to be similar for mee

would be about 700gb for 2x drizzle with 8k images

[merry perch]

maybe its that now siril produces these too

drizztmp files

so i think it may have gotten more space hungry in the stable release

[mental geyser]

oggi

[merry perch]

italian today

[stark ermine]

For your information, Starless model zenith is now in 1.1. I think it is now very good.

[olive hinge]

i will hav to check it out

[merry perch]

Will there be any way to not also remove noise in the starless image? Its a bit jarring I'd rather have a traditional star only removal, which I have tried and is very good indeed

[trim steppe]

still get that wonky line thats not processed on the left side with the new model

[olive hinge]

thats due to the way it breaks the image into tiles

i think its bc the side doesnt fit the ratio

so it leaves it out

[stark ermine]

Read the message in the console. At the beginning you have a comment to solve it.

[stark ermine]

Answer from SyQon:

I understand that the result can feel “too clean” compared to a traditional star removal, but this is a paradigm shift rather than excessive denoising.

The model does not apply an explicit noise-reduction step nor adaptive filtering. Instead, it operates as a latent signal estimator, optimized to maximize the effective SNR of the diffuse component. In simple terms, it does not remove noise — it reconstructs the signal that the noise was masking.

From a statistical standpoint, the model approximates a solution of the form:

Ŝ(x) = E[S(x) | I(x)]

where S(x) is the true diffuse signal and I(x) is the observed image. In this framework, noise is not something to be “preserved”: it is a stochastic, spatially incoherent variable that is automatically rejected because it does not contribute to a stable solution. This is why the noise can disappear almost completely without introducing blur or structural loss.

The key difference compared to traditional denoising methods is that those operate through local trade-offs (smoothing versus detail), while here the SNR increase happens globally, by exploiting spatial and multiscale coherence of the real signal. If a structure is supported by its context, it is reinforced; if it is not, it is suppressed. There are no arbitrary thresholds and no “strength” parameters.

It is also important to note that if noise is not present, the model has nothing to reject: in the absence of an incoherent stochastic component, the output converges almost exactly to the input. This is why it does not introduce new structures and does not artificially “enhance” already clean data.

In other words, the result is not “cleaner than reality”, but closer to the underlying physical signal than the noisy observation. Getting used to this type of output means finally separating the concept of “visible noise” from that of “useful information”, and working with a starless background that more accurately represents the astronomical content.

If the goal is to preserve the original noise texture for aesthetic reasons or compatibility with traditional workflows, then a classical star-only removal remains the appropriate choice. This model, instead, is designed to directly produce the best possible compromise between physical fidelity and readability, without requiring a subsequent denoising step.
It's not denoise, it's pure statistical math of the signal

[merry perch]

Interesting. If I were to pre-denoise it with Graxpert then, would it not find any high frequency noise to suppress?

Is this model made to work best on non-denoised images? Cause in my test with M45, no denoising, worked perfectly, with denoise it missed some big stars

[dark sleet]

I ran the comment but now SyQon apparently just disabled gpu acceleration by default

[rough urchin]

ngl i have a feeling that someone Is gonna upload starless files

[stark ermine]

The call is over

[olive hinge]

wha?

[zenith ridge]

Would be great to have an Android port of siril !

[stark ermine]

Haha.

I'm sorry, but that's something that will probably never happen.

First because GTK won't run on Android

Second because Android platforms are not really dedicated to image processing

[merry perch]

Lol, its actually manageable if you install a proot Linux distro on Termux and build from source. Honestly I chatgpt pasted my way through most of it but it worked

Some scripts may not work but I am surprised it runs at all

Thats the most Android running you'll ever do with Siril probably

[olive hinge]

@stark ermine during registration siril is forcing me to debayer my data, is there any way to make it stop?

im at 850nm so my bayer matrix is completely transparent

so it doesnt need debayering

all its doing is damaging the data

[stark ermine]

Remove bayer information in all your fit headers. (Siril can do it)

[olive hinge]

can i do that to all subs at once?

[merry perch]

Yes if you delete the keyword it'll do it to all subs of the sequence

[olive hinge]

how do you do it

[merry perch]

Go to the FITS headers, select the keyword and delete using the minus button i think

It'll give a warning telling you it will be done to all the sequence

[olive hinge]

wher is fits headers 😭

[merry perch]

Its in the Tools menu

[olive hinge]

a oki

[stark ermine]

yes

use the command line to apply it on all subs

https://siril.readthedocs.io/en/latest/Commands.html#seqheader

[subtle night]

Running or building SIRIL on Arch Linux currently fails due to the package llhttp, which should be downgraded to v9.2.1 for it to work again

[olive hinge]

breh

arch users am i right

[subtle night]

The worst human beings imaginable
-# I use arch btw

[olive hinge]

eeeew

use a real distro

boooo

[subtle night]

What is a real distro?

Ubuntu?

[olive hinge]

ew

god no

ubuntu is the shittiest shit ever shat onto this planet

[subtle night]

Hannah Montana Linux?

[olive hinge]

god yea, thats the linux distro of kings and queens

[subtle night]

What is a real distro though

and why

[olive hinge]

it was a joke

[subtle night]

Some people take that way too seriously, I wasn't sure lol

[olive hinge]

seriousness is not something i do often

[trim steppe]

windows:

[merry perch]

Stop distro wars

Obviously Temple OS is better

[subtle night]

I concede

[stark ermine]

this is not a direct dep of Siril

[subtle night]

@stark ermine is there any way to use filters of the "wrong" color for SPCC?

I see that Sloan irg can be mapped to the RGB channels, but I'd like to select Touptek R and G for my green and blue image channels, and add in my SVBONY SV183 longpass IR (685 nm) filter for the red channel.

[olive hinge]

i think you can just individually select the wavelength + bandpass

[subtle night]

Oh right I didn't even think about that

[merry perch]

I was wondering, what is the Noise weighting option really about? Just give more weight to less noisy frames? So if I were stacking 10s and 20s exposures together, the 20s would get more weight?

When do you think I should use that over wFWHM

[stark ermine]

Don't stack different exposures together

[merry perch]

Would it be better if I stacked them separately and then stacked the masters?

[stark ermine]

Nope

They must've combined in some HDR tools

[still sapphire]

What's the difference between Zenith v1 and axiom v2 on syqon starless?

[stark ermine]

They are different models

Zenith is free

[olive hinge]

so axiom v2 is paid?

[stark ermine]

yep

[olive hinge]

is it betterer?

[stark ermine]

I don't know 🙂

[olive hinge]

[olive hinge]

[merry perch]

Is this with AberrationRemover

Cause I have that too

I tried to reset the venv in the settings, but it didnt work

[olive hinge]

nop

with the star remove

[merry perch]

Try to use the new GPU helper script

See if there is anything to install like CUDA

If you have nvidia

Fixed it for me

[split smelt]

Does it matter if Zenith is good enough?

[olive hinge]

nah

[subtle night]

So I've got some frame data taken at different gain levels (gain 200 is what I should be using, but I have some data I took at gain 100) and I want to combine the gain 100 and gain 200 stacks while normalizing the values to that of the gain 200 stack, since I plan to do some continuum subtraction with more gain 200 data and combine it with IR and luminance data that was only taken at gain 200.

What's the best way of combining this data together and normalizing the intensities to match that of the gain 200 data?

[zenith ridge]

New cosmic clarity models from Frank with AI4

[zenith ridge]

I tested it in SAS Pro and the results are truly impressive, so it would be amazing to have it interfaced within Siril. Instead of running on the cosmic clarity executables, the script would need to be modified to call SAS Pro.

[stark ermine]

you need to contact the script owner, which is Adrian

[subtle night]

@stark ermine in the documentation for image normalization, what is meant by "scaled to achieve dispersion matching" in the methods that include scaling? Is it just multiplying the images after the background removal to match them more closely?

[stark ermine]

There's no background removal

Normalization tries to make images consistent between each others in order to use rejection algorithm

On a better way

With images having consistent statistics

[merry perch]

I have the same question specifically. Additive I seem to understand it means that it simply adds or subtracts from the background brightness, while scaling? Does it try to match the signal?

[stark ermine]

kind of yes

[subtle night]

*not background removal in the sense of using RBF/polynomial fitting, but adding/subtracting a constant to match overall brightness

that's what I meant by that

[stark ermine]

yep this is the additive part

[subtle night]

@stark ermine what is considered a rejected sample when Winsorized sigma clipping is used?

Just the samples that are rejected in the sigma clipping step, or the Winsorized samples as well?

[stark ermine]

If you are interesting in understanding the code, you can have a look here: https://gitlab.com/free-astro/siril/-/blob/master/src/stacking/rejection_float.c?ref_type=heads

[subtle night]

I wish there weren't as many magic numbers in the code, but from what I understand that's not on the SIRIL team

[stark ermine]

What do you mean??

What magic numbers??

[subtle night]

Where factors like 0.991 for the IKSS estimator come from, for instance. I see it in archives of the PixInsight ImageIntegration documentation

It's not clear how that factor is derived

[subtle night]

That, and the factor of 1.134 in the Winsorized sigma clipping algorithm: "the constant 1.134 is derived from the normal distribution for a sigma factor of 1.5 (as recommended in Huber's work)" and there's no further explanation of how this value is derived

[subtle night]

oh I found this: it's the cumulative probability of the data lying outside 1.5 sigma

So it should actually be 0.13361440253771617

0.13361440253771617 == 2 * (1 - (1/2 * (1 + erf(1.5 / sqrt(2))))) == 2 - (1 + erf(1.5 / sqrt(2)))

(well, you add 1 to that)

[subtle night]

After doing some numerical integration, this constant should be 0.9909048195090278

[stark ermine]

yep all that stuff come from statistics

there are no magic numbers

Just the assumption you have enough data and that follow the normal distribution

[subtle night]

"Magic" in the programming sense

Though some of these numbers aren't described very well in the source material

so I understand why that's the case

[subtle night]

I'm trying to stretch sequences in SIRIL for a video and I'm getting some really weird results - the stretch applied does not match the preview at all

My image gets nuked

This occurs with both the histogram stretch and GHS options

[subtle night]

What I expect (left) vs. what I get (right)

It also screws up the AutoStretch in the preview window unless I hit "clear sequence"

[brain@westerlund:~/astrophotography/projects/orion-2x2/2025-11-13/Orion Square Panel 3]$ diff pp_light_H_.seq pp_light_H_.seq.bak 
3c3
< S 'pp_light_H_' 1 42 42 5 0 6 0 0 0
---
> S 'pp_light_H_' 1 42 42 5 -1 6 0 0 0
47d46
< M0-0 9048064 9048063 0.00123745 0.00116188 0.000381033 -999999 7.07254e-05 -999999 -999999 -999999 0 0.248801 1 8.70506e-05

Here's a diff of what happens to the sequence file after I try stretching the sequence with GHS

[subtle night]

Opened an issue for it: https://gitlab.com/free-astro/siril/-/work_items/1934

[merry perch]

@stark ermine I don't know if this is a user issue, but here I have a pretty large number of subs that fail 2 pass registration on this dataset, some cloudy frames fail ofc, but also several good subs were rejected as well. I already tried some things, such as changing the reference frame before starting the first pass of registration; tweaked settings in the dynamic psf dialogue and also ticked "relax psf checks". Its not a massive number of subs, but I see some useful data being left behind epsecially considering there are worse frames being registered just fine. This only started to happen once I added 4-5 extra hours on this dataset. if you need anything else for an educated guess about the issue, i can provide it. Thanks.

[stark ermine]

If images aren't aligned, there must be a good reason. I regularly have hours of data to stack and never get any rejects. So you have to look at the logs to figure out what's causing the problem. You can also try aligning the images using plate-solving.

[merry perch]

Oh thats right i can use platesolving.

Well there is a batch of frames that is slightly shifted a bit compared to the others, but I thought there were enough stars to overlap

[stark ermine]

This is probably the issue

with astrometry it will work

[merry perch]

Also I was wondering. Since the astrometric registration works well on most datasets even with problematic overlaps, is there a reason not to always use it over global star alignment?

[stark ermine]

Users do want they want. There is no preferred choice

[subtle night]

The one thing I don't know about (and @stark ermine would be able to clarify) is whether undistortion corrections are done with astrometric registration alone (as opposed to global star alignment w/ undistortion from images)

[merry perch]

That would be interesting to know. Also, my real question was more if astrometric alignment is worth it 99% of the time over global star alignment

Cause I am starting to think that some times

Cause even without a mosaic I still run into occasions where subs taken close to each other still somewhat fail to register due to some reason, and mosaic registration usually fixes that

[stark ermine]

yes they are

[subtle night]

The only time I've had issues with astrometric registration is when there are very few stars and the distortion coefficients have a high uncertainty (e.g. shooting the Helix in narrowband with my RC)

[stark ermine]

You can lower the distortion coefficient

[subtle night]

I found this on the Astrobin forums...
https://www.youtube.com/watch?v=MXDgMY8Tyqk

[mental geyser]

thats pretty sick

[stark ermine]

Yep... That could be a game changer

of course, no need to sort the files

you can have a big mess, you drag&drop altogether

and tat's fine

[mental geyser]

something ive been wanting since 2022

[stark ermine]

it's almost done on my side

I hope

only for FITS images

necessary to have right FITS keywords

[floral plaza]

Gyatt damn

If only siril could support multiple files open in one place, it'd totally replace pix for me

[stark ermine]

you can do it with the multiwindow script

[safe shale]

For me the script feels a bit slow and doesn’t properly align the windows

And it’s not built in like pixinsight, where you can do a lot more manipulations, etc. (and they feel a lot easier and better to use).

[stark ermine]

Don't expect Siril to work like PixInsight

[safe shale]

Of course, and I don’t use pixinsight
It would be very nice if multiple windows was a built in feature

[merry perch]

Upon everything thats being added to Siril, i wonder (if its not against its core UI philosphy), if its going to be possible to pin favorite scripts somewhere easy to click instead of digging in the script menu every time

[steel grotto]

do old scripts stop working? i think this was originally a pix script

[stark ermine]

what do you mean?

old scripts always work

[steel grotto]

alright i did transfer checkbox and it was the one that reduces stars so still working

star checkbox just freezes

[stark ermine]

you can rise a ticket in our GitLab script repository

[deft meteor]

siril!

[stark ermine]

yes?

[deft meteor]

idk i was just saying siril

[deft meteor]

is siril vibe coded

[stark ermine]

Siril has been around since 2005... and we took over its code in 2012.

way before chatGPT

[trim steppe]

siril is that old???

[stark ermine]

Yes it is

http://dulle.free.fr/alidade/siril.php?priv=&lan=en

[deft meteor]

do you ensure no contributors use ai?

[stark ermine]

Not really. And it wouldn't be such a big deal either. AI can be useful. Why is that?

[deft meteor]

there was a huge convo about ai

[stark ermine]

AI is a hot topic.

[deft meteor]

ya

[merry perch]

My idea would be that as long as nobody is straight up vibe coding its fine to use AI as a tool under supervision

[stark ermine]

AI is great for debugging

[merry perch]

Sure is

Or repetitive tasks that is

The line for me is simply where the AI has to build the program for you

[split smelt]

AI is ideal for being a partner in pair programming, where you know what you're doing but want assistance https://en.wikipedia.org/wiki/Pair_programming

[deft meteor]

NO O

oop

my bad

did not mean to be in this chat or reply to you

[dark sleet]

before and after with SyQon starless ... Why is it applying wavelet transformation??

with starnet

[deft meteor]

syquon starless denoises images for some reason

well "denoises"

it is dumb

i dont use antthing from them due to this

[safe shale]

Same, it's annoying

[dark sleet]

:(

[hollow garnet]

in squares too lol

[dark sleet]

True lol

[merry perch]

Well yes. My honest feedback, is that for example, star removal itself is great, i just cant bring myself to use any of their stuff simply because it insists on performing things that then have nothing to do with the original purpose of the tool

[subtle night]

Love me some Unix philosophy: make one tool to do one task well, and then compose them

[stark ermine]

SyQon will update its starless model soon

without any denoiser