I am always afraid that Russ will come up with a tool called WorkXterminator that will essentially do all my work for me, and make me lazier than I already am. Nice work Russ and Adam!
Wow, great idea to have Russell on with you. I’d already played with the new version and been blown away, but now I’ll be able to experiment with the new sliders. Thanks for making this and posting it so quickly.
This is so good! it's been miserable weather here in Italy for quite some time, so no new images to work on. But who needs them when you can go back to all your images to experiment with this new tool? Thanks to both: Russ for creating it and Adam for giving us the chance to understand it better. As usual, I must add
This is amazing, thanks for bring this to us, Adam! And thanks for showing up to have this conversation, Russ! It really means a lot to hear from two subject matter experts. I used to use NXT on each mono channel before combination but it sounds like I’ll be saving NXT for later in the process now. Great improvement all around!
We knew Russ was busy while others were busy trying to replicate his results with free offerings for NR and Deconvolution. As good as some may have been, these latest improvements will definitely set the bar even higher for what's possible, and keep the X-suite far ahead of the competition! Even though BlurX gets most of the accolades, I shoot in B9 sky and NoiseX has always been my most valuable tool. This update is Fantastic!!
Excellent presentation, the instructions and examples were very clear. Seeing the noise being destroyed on the Iris Nebula had me looking like 😮 Thanks Russell for all the hard work you put into these tools and to Adam for putting this interview together. Looking forward to no doubt seeing updated walk throughs with Adam with narrowband processing 🙂
Glad that myth is finally busted about "sharpening the noise", and also the part about linear vs stretched data. The reason I initially purchased NoiseXT was after I spent a bunch of hours trying to use MLT and MMT and just not really being able to figure out exactly what noise was in which wavelet, and just going around in circles, so it's great that we can now do low and high frequency noise in NoiseXT. Really looking forward to checking this out..
Thanks! Nice to have the additional fine-tuning ability. Just when I feel I have a decent understanding of the noise issue, I realize I have more to learn! Well, maybe not so much more after watching this. The ability to selectively work with LF/HF noise is incredibly helpful.🙂
Outstanding. Thank you, gentlemen. (And what a treat, of the two examples, I have just captured both, just finished processing Iris, so I could follow along, and am about to process NGC1333 with the new AI.)
What I’m waiting for the most is an update to BlurX that could completely solve my tilt issue and the elongated stars in the corners. The current version already does an amazing job. I’ve bought every one of Russel’s programs, and I can’t imagine going back to the days when we had to solve those issues by hand. 😂
Simply amazing!! During the video I noticed the Real Time updated immediately without lag when demonstrating the different sliders. Why is that, as NXT is computationally heavy?
I was running NXT on an Apple silicon Mac... NXT (as most neural networks) is VERY computationally heavy, but Apple has done some great work with their GPUs.
very nice new functions. Been using lightness/colour noise slider in Lightroom for ever. Nice to have it here now. The frequency separation is nice to see too.
That was an awesome presentation, thank you so much for those new features, once again! I hope an updated BXT can handle my pinched hexagonal Esprit120 stars/spikes better as well as weird diffraction rings on bright stars.🤪🙏
Thanks for the video Adam and Russ. Looking forward to use the new version, PI did not alert me to it so I will have to make shows up (I just updated PI and it showed up). I like NXT a lot and use it mainly in the nonlinear state for a number reasons. I usually do a 50% Noise Reduction but then later in the processing I do it again at 40-50%. I wonder if that somehow has the same iterative effect that is on this new version. Cheers!
Thank you so much Russ and Adam! Very helpful in explaining the improvements to this powerful tool. Just in time for a project I am working on. Regarding workflow- if using all your tools on a linear image, is it best to run SX before NX? The image I am currently working on is OSC and 2X drizzled. Thanks so much!
Nova Scotia here. Hi all. Just updated my Noise Exterminator. Seems to have just been available in the last few minutes at about 1500GMT. Not sure I'll ever need the new flexibility as I am fully satisfied with the older version! So I'll likely just use the defaults 99% of the time.
Brilliant, have also just updated and find AI 3 very impressive 🙂 I gotta ask, is there any possibillity of a Flats exterminator being in the works ... ?
I don't have any plans. This is akin to gradient removal, and I've never been convinced that AI is the right approach. PI's MARS project is on the right track IMO.
Thank you for this presentation. Nice to see we have now such capability to filter differently colour and intensity! I would have also loved to hear from Russ why only nvidia can accelerate the processes. AMD GPU are so much cheaper, could they - one day - do the work too?
Thanks for the great interview! You talked about the sequence of *XT in the workflow, and it's pretty clear that BXT comes first. Currently I'm using NXT before SXT, but wonder whether it would also work to use NXT at a later stage, e.g. after combining starless RGB and Ha? Does SXT work better with oder without denoise before?
And regarding BXT (BlurXTerminator), did I get it right, using BXT is recommended on single channels BEFORE combining the channels? And in linear mode?
You can use BXT either before or after channel combination. In some cases it's better to do it after so that the neural net can see all three channels simultaneously... it can then match the final PSFs for each channel to one another. Just don't do any channel mixing prior to deconvolution - the blending of the channel PSFs will confuse the network, especially if one is gained up much more than the others, e.g., [S II] in a NB image. When doing NB, I like to do a straight SHO combination, then BXT, then whatever channel mixing is needed to get a palette I like.
@@rrcroman Thanks a lot! Regarding deconvolution, I dont do any deconvolution at all because I dont know what exactly it is and why I should use it. BXT and NXT offer very good results. I think deconvolution is outdated, isn´t it? (at least for deepsky pictures) KR
BXT is a form of deconvolution. Deconvolution is quite powerful in restoring contrast at fine scales. It's a form of sharpening, but a very mathematically formal one - it tries to reverse the blurring of an image due to atmospheric seeing and the telescope's optics.
@@rrcromanIf you are editing an OSC Dual NB image can you apply BlurXT without extracting the Ha and Oiii? I normally do that and then use Narrowband Normalization for an HOO. If I have two masterlights HA/OIII and SII/OIII then I use DBextract and run BlurXT after Channel Combination on the starless channels.
Please comment on the work flow between BXT and NXT. It was stated that NXT data should not be done after any deconvolution. Is BXT a deconvolution type function? I have seen Adam demonstrate BXT for small stars reduction in his Adam Block Studios Fundamentals video NGC 1491 workflow.
The correct statement is that noise reduction (of any kind, including NXT) should not be done BEFORE any form of deconvolution (BXT or otherwise). Deconvolution needs close to raw data. Calibration, integration, color calibration (e.g., SPCC) and gradient removal before deconvolution are OK, but not much else.
Looks great, but it's going to be hard to use a real time preview to get the settings right if you don't have CUDA. I wish there was an AMD way of getting CUDA to work, or similar.
Adam, not sure if you can answer this or if Russ is keeping an eye on comments here, but this is related to your short discussion about the tensorflow. I was using PI on Windows and had the tensorflow repository configured and it seemed to work as expected. I recently moved away from Windows and am using Kubuntu (PI's recommended OS). I have it setup some that it can take advantage of Cuda as well. I was wondering if Russ' tensorflow supersedes in Linux as it does in Windows. Thanks..... oh, and great video as usual!
@@rrcroman so there is no need for Linux users to include this repository? It's a shame as the tensorflow we have to use is rather "old". Anyway, thanks for updating the NXT tool and I look forward to whatever comes next.
@@thedeathtical If you add the TensorFlow repository on a Linux machine, it will do nothing. I hear you on the dated version of TF, but that's also true for Windows... the TF project stopped building the GPU version for Windows some versions ago. But the old version still works just fine.
I am always afraid that Russ will come up with a tool called WorkXterminator that will essentially do all my work for me, and make me lazier than I already am. Nice work Russ and Adam!
What an update to this already great tool. NoiseX, BlurX and StarX should be in everyone's work flow. Well done!
Wow, great idea to have Russell on with you. I’d already played with the new version and been blown away, but now I’ll be able to experiment with the new sliders. Thanks for making this and posting it so quickly.
I'm sure I speak for everyone here...thank you both so very much. What a time to be alive!
This is so good! it's been miserable weather here in Italy for quite some time, so no new images to work on. But who needs them when you can go back to all your images to experiment with this new tool? Thanks to both: Russ for creating it and Adam for giving us the chance to understand it better. As usual, I must add
This is amazing, thanks for bring this to us, Adam! And thanks for showing up to have this conversation, Russ! It really means a lot to hear from two subject matter experts. I used to use NXT on each mono channel before combination but it sounds like I’ll be saving NXT for later in the process now. Great improvement all around!
Thank you both so much for taking the time to show how powerful these tools continue to grow!
Amazing! Can't wait to play with it. Thanks Adam and Russ!!
We knew Russ was busy while others were busy trying to replicate his results with free offerings for NR and Deconvolution. As good as some may have been, these latest improvements will definitely set the bar even higher for what's possible, and keep the X-suite far ahead of the competition!
Even though BlurX gets most of the accolades, I shoot in B9 sky and NoiseX has always been my most valuable tool. This update is Fantastic!!
Excellent! Many thanks Russ and Adam
Wow! Thanks Russ!!! Thanks Adam! I'm going to be busy today ;-)
Excellent presentation, the instructions and examples were very clear. Seeing the noise being destroyed on the Iris Nebula had me looking like 😮 Thanks Russell for all the hard work you put into these tools and to Adam for putting this interview together. Looking forward to no doubt seeing updated walk throughs with Adam with narrowband processing 🙂
Awesome. Thank you both 🙏🏻💞
Glad that myth is finally busted about "sharpening the noise", and also the part about linear vs stretched data.
The reason I initially purchased NoiseXT was after I spent a bunch of hours trying to use MLT and MMT and just not really being able to figure out exactly what noise was in which wavelet, and just going around in circles, so it's great that we can now do low and high frequency noise in NoiseXT.
Really looking forward to checking this out..
Amazing job Russ and thank you Adam!
Thanks! Nice to have the additional fine-tuning ability. Just when I feel I have a decent understanding of the noise issue, I realize I have more to learn! Well, maybe not so much more after watching this. The ability to selectively work with LF/HF noise is incredibly helpful.🙂
Thanks for this, and for highlighting the most salient moments as you did. Very clear presentation.
Only 20 days till I get back and try this. So pumped. Edit: I also love seeing how he prefers to use it
Outstanding. Thank you, gentlemen. (And what a treat, of the two examples, I have just captured both, just finished processing Iris, so I could follow along, and am about to process NGC1333 with the new AI.)
I have this and am very excited to try these new versions. Great video. Very helpful and engaging. Thankyou both very much!
Great presentation! Thanks for all you and Adam do for the Astro community.
Phenomenal! Thanks for bringing this to us.
Works perfect ! Superb !
Thanks for this. Much appreciated. Most in depth analysis available bar none. RC’s tools have transformed our hobby. Cheers, Des
What I’m waiting for the most is an update to BlurX that could completely solve my tilt issue and the elongated stars in the corners. The current version already does an amazing job. I’ve bought every one of Russel’s programs, and I can’t imagine going back to the days when we had to solve those issues by hand. 😂
Thanks for the wonderful work and presentation!
Thank you guys, this was an incredible learning experience!
Simply amazing!! During the video I noticed the Real Time updated immediately without lag when demonstrating the different sliders. Why is that, as NXT is computationally heavy?
I was running NXT on an Apple silicon Mac... NXT (as most neural networks) is VERY computationally heavy, but Apple has done some great work with their GPUs.
Amazing!! 🎉 Thankyou
very nice new functions. Been using lightness/colour noise slider in Lightroom for ever. Nice to have it here now. The frequency separation is nice to see too.
Amazing Thank You verry much
My hero!!
Wow amazing 🤩
Amazing. Thank you very much to both of you. You are stars :-)
I hope the new version of star xterminator can better differentiate between stars and galaxies. 🤞can't wait.
That's one of the primary goals of the next version. 🙂
That was an awesome presentation, thank you so much for those new features, once again!
I hope an updated BXT can handle my pinched hexagonal Esprit120 stars/spikes better as well as weird diffraction rings on bright stars.🤪🙏
Totally brilliant.
Thanks for the video Adam and Russ. Looking forward to use the new version, PI did not alert me to it so I will have to make shows up (I just updated PI and it showed up). I like NXT a lot and use it mainly in the nonlinear state for a number reasons. I usually do a 50% Noise Reduction but then later in the processing I do it again at 40-50%. I wonder if that somehow has the same iterative effect that is on this new version. Cheers!
Perhaps, but NXT AI2 was a bit prone to creating false stars if run iteratively. AI3 solves that.
Thank you so much Russ and Adam! Very helpful in explaining the improvements to this powerful tool. Just in time for a project I am working on. Regarding workflow- if using all your tools on a linear image, is it best to run SX before NX? The image I am currently working on is OSC and 2X drizzled. Thanks so much!
You can take your pick on the order in which to use NXT and SXT - it doesn't make much difference.
Great video and look forward to more new features on there way, for the rest of the Xterminator family 👍🏻
awesome!!!!!
Nova Scotia here. Hi all. Just updated my Noise Exterminator. Seems to have just been available in the last few minutes at about 1500GMT.
Not sure I'll ever need the new flexibility as I am fully satisfied with the older version! So I'll likely just use the defaults 99% of the time.
Brilliant, have also just updated and find AI 3 very impressive 🙂
I gotta ask, is there any possibillity of a Flats exterminator being in the works ... ?
I don't have any plans. This is akin to gradient removal, and I've never been convinced that AI is the right approach. PI's MARS project is on the right track IMO.
When AI 4.0 comes out, we won't need no frickin' telescopes... ;)
Amazing. But now I need to reprocess everything I've ever captured again! 😆😆😆
If it makes you feel any better, so do I. 🙂
Thank you for this presentation. Nice to see we have now such capability to filter differently colour and intensity! I would have also loved to hear from Russ why only nvidia can accelerate the processes. AMD GPU are so much cheaper, could they - one day - do the work too?
Good morning from Indiana
Good job guys!
Thanks for the great interview! You talked about the sequence of *XT in the workflow, and it's pretty clear that BXT comes first. Currently I'm using NXT before SXT, but wonder whether it would also work to use NXT at a later stage, e.g. after combining starless RGB and Ha? Does SXT work better with oder without denoise before?
And regarding BXT (BlurXTerminator), did I get it right, using BXT is recommended on single channels BEFORE combining the channels?
And in linear mode?
You can use BXT either before or after channel combination. In some cases it's better to do it after so that the neural net can see all three channels simultaneously... it can then match the final PSFs for each channel to one another. Just don't do any channel mixing prior to deconvolution - the blending of the channel PSFs will confuse the network, especially if one is gained up much more than the others, e.g., [S II] in a NB image. When doing NB, I like to do a straight SHO combination, then BXT, then whatever channel mixing is needed to get a palette I like.
@@rrcroman Thanks a lot! Regarding deconvolution, I dont do any deconvolution at all because I dont know what exactly it is and why I should use it. BXT and NXT offer very good results. I think deconvolution is outdated, isn´t it? (at least for deepsky pictures) KR
BXT is a form of deconvolution. Deconvolution is quite powerful in restoring contrast at fine scales. It's a form of sharpening, but a very mathematically formal one - it tries to reverse the blurring of an image due to atmospheric seeing and the telescope's optics.
@@rrcromanIf you are editing an OSC Dual NB image can you apply BlurXT without extracting the Ha and Oiii? I normally do that and then use Narrowband Normalization for an HOO. If I have two masterlights HA/OIII and SII/OIII then I use DBextract and run BlurXT after Channel Combination on the starless channels.
Please comment on the work flow between BXT and NXT. It was stated that NXT data should not be done after any deconvolution. Is BXT a deconvolution type function? I have seen Adam demonstrate BXT for small stars reduction in his Adam Block Studios Fundamentals video NGC 1491 workflow.
The correct statement is that noise reduction (of any kind, including NXT) should not be done BEFORE any form of deconvolution (BXT or otherwise). Deconvolution needs close to raw data. Calibration, integration, color calibration (e.g., SPCC) and gradient removal before deconvolution are OK, but not much else.
Is this update for the Photoshop version as well?
Not yet... I'm working on it right now. 🙂
@ thanks, Russ!
Looks great, but it's going to be hard to use a real time preview to get the settings right if you don't have CUDA. I wish there was an AMD way of getting CUDA to work, or similar.
I'm looking into enabling other brands of GPUs, under Windows at least. Don't hold your breath... it remains to be seen if it's possible.
Adam, not sure if you can answer this or if Russ is keeping an eye on comments here, but this is related to your short discussion about the tensorflow. I was using PI on Windows and had the tensorflow repository configured and it seemed to work as expected. I recently moved away from Windows and am using Kubuntu (PI's recommended OS). I have it setup some that it can take advantage of Cuda as well. I was wondering if Russ' tensorflow supersedes in Linux as it does in Windows. Thanks..... oh, and great video as usual!
The TensorFlow repository mentioned in the video only applies to Windows. Linux users still have to set things up manually as you've done.
@@rrcroman so there is no need for Linux users to include this repository? It's a shame as the tensorflow we have to use is rather "old". Anyway, thanks for updating the NXT tool and I look forward to whatever comes next.
@@thedeathtical If you add the TensorFlow repository on a Linux machine, it will do nothing. I hear you on the dated version of TF, but that's also true for Windows... the TF project stopped building the GPU version for Windows some versions ago. But the old version still works just fine.
Is there any exeleration (multi core) options for AMD video cards?
Not yet, but I'm looking into it.