mcbroomf wrote:
I've found this as well and came up with a simple method to do c) ie redo the photo.
Throw out the DNG
Make a virtual copy of the processed RAW file
Reset the virtual file (ie drop the post processing)
Apply Denoise to the virtual file you just reset
Copy the post processing from the processed raw file to the denoised DNG
Throw out the virtual copy
It may seem a little complicated to describe but it only takes a few moments (except to run another denoise)
I prefer to work on my raw files to see if they are keepers, then run denoise on those that survive. That way I have much less time waiting for denoise to run (30 secs or so for each file). Most don't seem to be affected but for those that are I gave up trying to find a common root cause and just use the steps above.
Mike, thanks for posting this. It looks like a good solution and I will give it a try.
Once I have finished my culling, my next step is to apply cropping (only) to the remaining photos, as this gives me an even better feel for a photo's potential and allows me to drop some more photos from consideration before I get into all the rest of the post-processing. On a rare occasion I will complete processing of a "surviving" photo and then decide I want to crop it even more. This is where I run into trouble (as is the case with my aforementioned photo).
mcbroomf wrote:
I've found this as well and came up with a simple method to do c) ie redo the photo.
Throw out the DNG
Make a virtual copy of the processed RAW file
Reset the virtual file (ie drop the post processing)
Apply Denoise to the virtual file you just reset
Copy the post processing from the processed raw file to the denoised DNG
Throw out the virtual copy
It may seem a little complicated to describe but it only takes a few moments (except to run another denoise)
I prefer to work on my raw files to see if they are keepers, then run denoise on those that survive. That way I have much less time waiting for denoise to run (30 secs or so for each file). Most don't seem to be affected but for those that are I gave up trying to find a common root cause and just use the steps above.
What is the difference between your steps and what LR does when DeNoise is used?
My understanding is that LR applies DeNoise to the unprocessed RAW data to produce a DNG file, then applies to the DNG file any edits that may have been applied to the RAW file.
I use Topaz... compared the two and both are very good.
I didn't read here about performance. With Adobe, you need a modern, expensive graphics card. Topaz does its thing in a fraction of the time with older cards. If you planning on buying a new card to use Adobe and all ready have Topaz... save your money, you are not getting anything more.
Besides, I guess I am a bit of an outlier, but I like Topaz more.
dclark wrote:
What is the difference between your steps and what LR does when DeNoise is used?
My understanding is that LR applies DeNoise to the unprocessed RAW data to produce a DNG file, then applies to the DNG file any edits that may have been applied to the RAW file.
It seems to vary. Masks especially can change. Just working on some Astro files with 2 masks and getting big blue cast on the first dng.
Zenon Char wrote:
I love it and I don't care if others edge it out. I've watched many videos about it and comparisons to other AI apps. Some presenters showed across the same file areas of Adobe Denoise AI did better while DXO did better in other areas.
...
20000 ISO with a pretty hefty crop. Next one is 16000.
Just asking: is it really ISO 20,000 or did you mean ISO 2,000?
So does this mean I can do milky way shots with an f4 lens like the 14-35? I've been trying to find f2.8 or faster extra wide primes for a couple of years without success. They are either extremely large and heavy or expensive, or don't take filters or all of the above. Just asking.
Jeff Nolten wrote:
So does this mean I can do milky way shots with an f4 lens like the 14-35? I've been trying to find f2.8 or faster extra wide primes for a couple of years without success. They are either extremely large and heavy or expensive, or don't take filters or all of the above. Just asking.
I'm not sure who sponsors him. He mentions Adobe not looking sharp but he has sharpening turned off when he applies Denoise. I do that too and add it back later. I also kept Topaz Sharpen AI in case files need a little more.
From the comments
Agreed, I tried it on some of my atro images from my f/4 lense and it’s impressive. It’s so good I can’t convince myself to upgrade to a faster lense. I can’t afford one anyways but now it’s not as big of a deal. Super impressed, thank you Lightroom and adobe!
Bruce n Philly wrote:
I use Topaz... compared the two and both are very good.
I didn't read here about performance. With Adobe, you need a modern, expensive graphics card. Topaz does its thing in a fraction of the time with older cards. If you planning on buying a new card to use Adobe and all ready have Topaz... save your money, you are not getting anything more.
Besides, I guess I am a bit of an outlier, but I like Topaz more.
pr4photos wrote:
I think I need a mac mini with an M processor. My 2018 3.2 GHz 6-Core Intel Core i7 takes about 8mins to process a Nikon Z7 45mp image with denoise!
It is not the CPU but the GPU that makes a difference for AI-denoise.
My Z7II files are denoised in 13 sec with a RTX 4000 Ada Generation laptop GPU.
Kaj E wrote:
It is not the CPU but the GPU that makes a difference for AI-denoise.
My Z7II files are denoised in 13 sec with a RTX 4000 Ada Generation laptop GPU.
My OM-1 files are just over 20MB and my M2 Max Mac Studio processes files in 8 to 11 seconds. The 2014 iMac took almost 2 minutes. As more features like Enhance become available we'll be uploading files to a Cray for processing. It still seems incredible to me that I now can get useful files (web, PDFs, family sharing) at ISO 25000 to 40000 with a m43 camera. I've made some decent prints as well. I've also rescued old Canon files when ISO 1600 was useless. I've learned never to delete a RAW file; I stack it with the Enhanced one. Never know when something even better will come along.
Even got some shots of bats at twilight; not really proud of them but amazed that the camera even focused at that light level, and at ISO 40,000.
I use M1 Max (Mac Studio) and the speed is satisfactory for Topaz and PS/LR.
My PC has RTX 4090 that should be substantially faster (that machine is noisy and power hungry) but I haven't used it for photo editing for a good while (LR catalog changed to Mac file path). My raw files are mostly in a NAS, although with 10Gbe ethernet to both Mac and PC, I'm not demanding ultimate speed. NR in LR with 71MB A7R5 Raw takes ~33 seconds (produced a 115.6MB DNG), I don't put raw files to my internal SSD (I have a TB4 external SSD but that's for the catalog or maybe videos not for raw photos) so I haven't tested.
IMO Topaz is better when it works the problem is Topaz is very inconsistent (and not very reliable), there are times it introduce unacceptable artifacts and for I don't know how long (at least for a few minor released versions) it messes up colour of A7R4 and A7R5 files (original raw file is intact, but colour of processed DNG is not good, interesting, raw from my older cameras are fine).
