As happens with beta software, there's a bug in the Stars-only deconvolution, reported by some of you, and confirmed with the devs. When you run the Stars-only deconvolution, and then run the denoise, it loses the stars-only deconvolution. The work around is to perform the Stars-only deconvolution, save the result, load that result file into GraXpert, and then perform the denoise. The bug should be fixed in a future release.
I've been holding off of buying Pixinsight because I really like working in siril/graxpert. Eagerly awaiting version 1.4 for drizzle integration. Only thing that keeps me thinking about Pixinsight is blur-x. Eventually there will be a siril tool that hopefully rivals that tool.
Just some feedback. I wrote a Windows C# GUI wrapper for my own use for the GraXpert 3.1.0rc2 version CLI (the supplied one crashes occasionally here). It automates the re-use of previously deconvolved results and is useful until it is no longer necessary. So far the GraXpert CLI has performed flawlessly with my GUI. I really love GraXpert !!!
I'm quite excited about this, but the 3.1.Orc2 beta hasn't been working very well in my trials so far. The star deconvolution seems very extreme - it seemed too much to me at anything stronger than 0.1; and more critically it sometimes makes medium brightness stars into bright green (at any strength). The object deconvolution, on the other hand, seems to have very little effect at low strength values, and at anything over about 0.7 sometimes introduces random coloured dots into the image. I have been using the method suggested in this video to get FWHM values in Siril. Also, as noted, you have to keep saving and coming back in so as not to lose one process when you apply another (though that wouldn't stop me using the processes, if they worked well). I'm also sticking with the version 2 AI denoise, as the version 3 runs very slowly, and sometimes produces very obvious nasty artefacts (e.g. obvious coloured rectangles). Images from the Seestar S50 seem particularly prone to these problems.
Saw the note about the bug on losing results but will the work process be to run the object decon against the gradient corrected image, and then the star decon on the result of that process, then denoise and save? Sorry for the confusion.
At the minute, I seem to be getting better results with the seti astro suite, although I believe he shares his knowledge with graxpert , perhaps it will get better shortly
@@DeepSpaceAstro Thank you!! After stacking in Siril, do you photometric color calibrate between background extraction, deconvolving or denoising in Graxpert?
I've been stacking and then crop in Siril. You can crop in GraXpert as well, it's just my preference. I save the cropped stack, open it in GraXpert, run BGE, decon, denoise, and then save. Then back in Siril for PCC.
hi there ! so for now (1.25) i can'T just go into preferences/miscellaneous and add graxpert there ? but in 1.4 i will be able to ? for now i need to use graxpert standalone ? tyvm
Really enjoy your tutorials. When loading the image into Grexpert, it flips it vertically, right? Also, after running Deconvolution, even on strength 0.3, I get some green stars. So, I think I'll try removing the stars with Starnet and deconvoluting just the star mask, then desaturate to see if I can tame those green stars without affecting the saturation of the nebula.
Yeah the flip is how GraXpert determines the data in the header. They do it differently than some other programs. Could you send me the unprocessed stack so I can take a look?
@@DeepSpaceAstro Thanks for your response Rich.Yes i can see some tabs: Conversion-Sequence-Callibration and Registration but no Console. Im running Ver1.2.4. Thanks Again.
@@DeepSpaceAstro On Affinity Photo, I think I have sussed it, it was the saturation setting that ws incorrect, just did two images both appeared normally...
i cannot download the stellar ai model. does anybody have any suggestions? tried deleting the old and redownloading the new with Anti virus turned off but no luck.
Awesome! Thanks! I will apply to last nights imaging today and weave into the workflow. Have you done a video on Siril's Drizzle? I'm wondering if worth doing. I read that Siril's drizzle isn't the same is pixinsight's and it's not very good.
Currently Siril's drizzle function is nothing more than a 2x upscale. The new version 1.4, that should be released early this year, will have true drizzle functionality.
Thanks for sharing this update. I ran the stars-only deconvolution and ended up with purple fringed stars. I had to reduce the strength all the way down to 0.15 to get reasonable results. I wonder if I had to use the object only deconvolution before the stars-only. I performed the by object-only deconvolution beforehand. Should I go the other way around?
@@DeepSpaceAstro Correct Rick. As you pointed out in your email, the FWHM setting was too high for my particular image. I had to reduce the value, and the deconvolution worked fine. Something to keep in mind, folks.
Separate to the issue I have about image saving - does anyone else have an issue with GraXpert crashing ? In Windows 10 GraXpert here about half the time locks up (not responding). I regularly have to go into Task Manager to end the main task. It also regularly leaves 'orphan' processes (further down the list in Task Manage) which have to be ended before re-starting GraXpert. This also happened in previous versions. P.S. I tried with and without AI hardware acceleration - no difference.
@@DeepSpaceAstro After looking in the log file when GraXpert crashes it mentions pipes and references to python files. Looking online shows many references to the 'pipe problem' with other application development. On a hunch I updated Python to the latest version (3.13.1). I don't know whether it is pure coincidence - but I've just done all the processing steps and twiddled sliders on two separate objects and GraXpert didn't crash. Fingers crossed that this has fixed my crash problems (although I've probably just jinxed myself...)
@@DeepSpaceAstro Unfortunately - although crashes seem less frequent - I still have them. Unless the command-line version behaves differently, when GraXpert is integrated with SIRIL any crashes may become more costly by hanging SIRIL (?). It is sad for me as GraXpert is amazing - apart from the crashes.
I saved after denoising and it just saved the denoising, and threw out the star deconvolution. I had to reload the original file, perform star deconvolution, save that, quit GraXpert, open it and load the deconvoluted file then denoise it in order to have an image that was star deconvolved plus denoised. Has anyone else experienced that?
@@DeepSpaceAstro I also have exactly this same issue. After deconvolution, I apply denoising to the stars only deconvolved image (selected). The denoising appears to undo the sharpening done by the deconvolution. If I save the deconvolved image I get the deconvolved image as expected. But - if I save the denoised image I get the 'un-deconvolved' denoised image. I cannot get the combined deconvolved/denoised image except by doing the steps given by PantherBlitz.
@@DeepSpaceAstro Mmmm - something strange here. There appears to be a disconnect between what's seen on the screen and what's saved. I'll do a series of screenshots and put into a PDF to investigate further.
@@DeepSpaceAstro I just tried with a different file. I can flip between original, deconvolved, and denoised and see that it is denoising the original while leaving the deconvolved untouched. Thus, the only way that I can get it to denoise the deconvolved image is to save it and load it again. I'm pretty sure that previous versions of GraXpert would incorporate all of previous steps into the final image. Not really a deal-breaker once one is aware of what is going on.
As happens with beta software, there's a bug in the Stars-only deconvolution, reported by some of you, and confirmed with the devs.
When you run the Stars-only deconvolution, and then run the denoise, it loses the stars-only deconvolution. The work around is to perform the Stars-only deconvolution, save the result, load that result file into GraXpert, and then perform the denoise.
The bug should be fixed in a future release.
Also, if you run the star decon and then run the object decon you loose the star decon. They go back to their original shapes and size.
Oh wow, GraXpert being integrated into Siril is gonna be awesome
Another great video. It's amazing watching these free tools evolve and improve over the years, thanks for keeping us up to date as usual.
Thanks! Yep they really help the community!
Awesome, so nice to get these updates and instructions Rich, much appreciated.
Very welcome! Thanks!
Happy New Year! As always, timely, very well put and to the point. Much appreciated, Rich!
Happy New Year to you too! Glad it was helpful!
As always, great video Rich!! Thank you so much for all the effort you put into the your channel and vids
Thanks for the kind words!
So much good stuff coming down the pipeline - and all for free! What a great time to be getting into astrophotography.
It really is!
It really is important to donate a few bucks when you can to show your appreciation of their efforts...
Great video, Richard. I love how this tool is improving!
Thanks! Yeah it really it is!
Excellent new feature! Thanks for demoing it Rich!
Good stuff!
This is so exciting! Thanks for update, Rich :)
Great Video as always. Thanks for the chance of learning 😇✌️
Thanks! Glad it was helpful!
Cool, thanks for the explanation and demo!
Glad it was helpful!
I've been holding off of buying Pixinsight because I really like working in siril/graxpert. Eagerly awaiting version 1.4 for drizzle integration. Only thing that keeps me thinking about Pixinsight is blur-x. Eventually there will be a siril tool that hopefully rivals that tool.
100% agree
Can't wait for 1.4 to get here
Me too!
Just some feedback. I wrote a Windows C# GUI wrapper for my own use for the GraXpert 3.1.0rc2 version CLI (the supplied one crashes occasionally here). It automates the re-use of previously deconvolved results and is useful until it is no longer necessary. So far the GraXpert CLI has performed flawlessly with my GUI. I really love GraXpert !!!
Awesome! Since you're a programmer, maybe you could also contribute to Siril as well!
@@DeepSpaceAstro Unlikely - my programming skills are on par with my astrophotography skills - which are both best described as 'hobby-level'.
I'm quite excited about this, but the 3.1.Orc2 beta hasn't been working very well in my trials so far. The star deconvolution seems very extreme - it seemed too much to me at anything stronger than 0.1; and more critically it sometimes makes medium brightness stars into bright green (at any strength). The object deconvolution, on the other hand, seems to have very little effect at low strength values, and at anything over about 0.7 sometimes introduces random coloured dots into the image. I have been using the method suggested in this video to get FWHM values in Siril. Also, as noted, you have to keep saving and coming back in so as not to lose one process when you apply another (though that wouldn't stop me using the processes, if they worked well). I'm also sticking with the version 2 AI denoise, as the version 3 runs very slowly, and sometimes produces very obvious nasty artefacts (e.g. obvious coloured rectangles). Images from the Seestar S50 seem particularly prone to these problems.
Saw the note about the bug on losing results but will the work process be to run the object decon against the gradient corrected image, and then the star decon on the result of that process, then denoise and save? Sorry for the confusion.
Before you denoise, save then open that saved file, then denoise.
At the minute, I seem to be getting better results with the seti astro suite, although I believe he shares his knowledge with graxpert , perhaps it will get better shortly
Multiple tools are always a good thing.
Great video, as usual 🙂
At what point of the workflow is it better to use graxpert deconvolution?
Thanks! The data should be linear, so before stretching.
@@DeepSpaceAstro Thank you!! After stacking in Siril, do you photometric color calibrate between background extraction, deconvolving or denoising in Graxpert?
I've been stacking and then crop in Siril. You can crop in GraXpert as well, it's just my preference. I save the cropped stack, open it in GraXpert, run BGE, decon, denoise, and then save. Then back in Siril for PCC.
hi there ! so for now (1.25) i can'T just go into preferences/miscellaneous and add graxpert there ? but in 1.4 i will be able to ? for now i need to use graxpert standalone ? tyvm
That's correct.
Another graet vidoe thanks for keeping it simple
Thanks!
Really enjoy your tutorials. When loading the image into Grexpert, it flips it vertically, right? Also, after running Deconvolution, even on strength 0.3, I get some green stars. So, I think I'll try removing the stars with Starnet and deconvoluting just the star mask, then desaturate to see if I can tame those green stars without affecting the saturation of the nebula.
Yeah the flip is how GraXpert determines the data in the header. They do it differently than some other programs. Could you send me the unprocessed stack so I can take a look?
Hiya Rich. I don't have the Console option , do i need to do something to add it. Great work by the way.
Assuming you don't see any of the tabs, there's an arrow at the center-right of the screen. Click that and it'll expand out.
@@DeepSpaceAstro Thanks for your response Rich.Yes i can see some tabs: Conversion-Sequence-Callibration and Registration but no Console. Im running Ver1.2.4. Thanks Again.
What if you hover over the vertical line between those tabs and your image? Can you click and drag to the left?
@@DeepSpaceAstro Rich your a Star.That did the job thank you so much.
Tried this put only displays my colour images in monochrome? (appear colour in Graxpert)
Where is it only displaying in mono, if it appears color in GraXpert?
@@DeepSpaceAstro On Affinity Photo, I think I have sussed it, it was the saturation setting that ws incorrect, just did two images both appeared normally...
i cannot download the stellar ai model. does anybody have any suggestions? tried deleting the old and redownloading the new with Anti virus turned off but no luck.
You can try to download and install them manually. github.com/Dark-Matters-Astro/graxpert-ai-models
Awesome! Thanks! I will apply to last nights imaging today and weave into the workflow. Have you done a video on Siril's Drizzle? I'm wondering if worth doing. I read that Siril's drizzle isn't the same is pixinsight's and it's not very good.
Currently Siril's drizzle function is nothing more than a 2x upscale. The new version 1.4, that should be released early this year, will have true drizzle functionality.
Hi, is it possible to download Siril with the integration of GraXpert? Thanks - PS: Like always great video!!
Thanks! It's still in development at the moment. They've said early this year for the beta release.
Siril update with GraXpert Cominges soon??
THX for the info
Yep, they're working on it. They said early this year.
Thanks for sharing this update. I ran the stars-only deconvolution and ended up with purple fringed stars. I had to reduce the strength all the way down to 0.15 to get reasonable results. I wonder if I had to use the object only deconvolution before the stars-only. I performed the by object-only deconvolution beforehand. Should I go the other way around?
I don't think it matters. Could you send me the unprocessed stack so I can take a look?
@@DeepSpaceAstro Sure I just sent you an email with a link. Thanks a lot!
@@DeepSpaceAstro Issue solved. Thank you Rich. You are the man!
@@guyyanez6949 For others that may see this...Was it you had your FWHM set too high?
@@DeepSpaceAstro Correct Rick. As you pointed out in your email, the FWHM setting was too high for my particular image. I had to reduce the value, and the deconvolution worked fine. Something to keep in mind, folks.
Separate to the issue I have about image saving - does anyone else have an issue with GraXpert crashing ? In Windows 10 GraXpert here about half the time locks up (not responding). I regularly have to go into Task Manager to end the main task. It also regularly leaves 'orphan' processes (further down the list in Task Manage) which have to be ended before re-starting GraXpert. This also happened in previous versions. P.S. I tried with and without AI hardware acceleration - no difference.
I've also had this issue on Win11, but I don't have an answer on why it happens.
I wonder if the developers are aware ?
@@DeepSpaceAstro After looking in the log file when GraXpert crashes it mentions pipes and references to python files. Looking online shows many references to the 'pipe problem' with other application development. On a hunch I updated Python to the latest version (3.13.1). I don't know whether it is pure coincidence - but I've just done all the processing steps and twiddled sliders on two separate objects and GraXpert didn't crash. Fingers crossed that this has fixed my crash problems (although I've probably just jinxed myself...)
@@DeepSpaceAstro Unfortunately - although crashes seem less frequent - I still have them. Unless the command-line version behaves differently, when GraXpert is integrated with SIRIL any crashes may become more costly by hanging SIRIL (?). It is sad for me as GraXpert is amazing - apart from the crashes.
Do I need to redownload graXpert or do I update it somehow?
Yes you need to download the latest version. Link is in the description.
@DeepSpaceAstro thankyou ❤️🙏
I saved after denoising and it just saved the denoising, and threw out the star deconvolution. I had to reload the original file, perform star deconvolution, save that, quit GraXpert, open it and load the deconvoluted file then denoise it in order to have an image that was star deconvolved plus denoised. Has anyone else experienced that?
You just need to be careful to select the correct image to save. It is possible to select any of them.
I haven't experienced that. Is it possible you had the wrong image selected when saving?
@@DeepSpaceAstro I also have exactly this same issue. After deconvolution, I apply denoising to the stars only deconvolved image (selected). The denoising appears to undo the sharpening done by the deconvolution. If I save the deconvolved image I get the deconvolved image as expected. But - if I save the denoised image I get the 'un-deconvolved' denoised image. I cannot get the combined deconvolved/denoised image except by doing the steps given by PantherBlitz.
@@DeepSpaceAstro Mmmm - something strange here. There appears to be a disconnect between what's seen on the screen and what's saved. I'll do a series of screenshots and put into a PDF to investigate further.
@@DeepSpaceAstro I just tried with a different file. I can flip between original, deconvolved, and denoised and see that it is denoising the original while leaving the deconvolved untouched. Thus, the only way that I can get it to denoise the deconvolved image is to save it and load it again. I'm pretty sure that previous versions of GraXpert would incorporate all of previous steps into the final image. Not really a deal-breaker once one is aware of what is going on.
The man 😎
💪
one question: when new siril version? :(
Early this year.
Does it fix coma ?
Not sure. Best I can say if give it a try and let us know.
Yes yes. I am giving it a try. But I am facing prblm of log while installing the model. Idk why
With all the time and energy it takes you to make these videos, how hard could it be for viewers to hit the thumbs up…really!! Come on people.
Yeah, and subscribe and share! 🤪 It's all good! Thanks!
Waiting Siril 1.4 final.
They're saying early this year!