Everything You Need to Know About Denoise in Lightroom Classic 

AMEN

Julieanne is the definitive source direct from Adobe!

Whoa! What GPU are you using? With my 3080 (and AMD 5800x cpu) it takes about 5 seconds.

Julieanne Kost does a great job explaining how this AI Denoise works but does not say "where" it happens. My wild-assed guess is that it happens in Adobe Camera Raw (ACR). Lightroom Classic (LrC), like Photoshop (Ps), cannot "raw process" and both run the ACR plug-in at that. Most people are not aware of this in LrC because it is masked in the User Interface (UI). But in its website, Adobe clearly indicates using the ACR plug-in in LrC. That is relevant to your question, because both the luminance and colour noise you see in example images, in my opinion, can be considered as artefacts following from raw processing. It is conceivable that the colour and luminance noise is generated in a similar way that Moiré is generated by raw processing. That is hard to prove and I would counter such a reply to my point by saying, well then it is insufficient deBayerisation (the first main part of raw processing) and demosaicking (the second part in raw processing). The Denoise algorithm we see at work in the video does a pretty good job. But detail retention, or prevention of detail loss, probably can be improved upon - and sharpening is another ACR weakness - very bad. Together with detail retention/sharpness another thing that can be a lot better is the upsampling to much higher MP images. So, if Denoise happens in ACR then you have it in Ps as well. Note here that I am subscribed to the Adobe photography bundle for over 10 years. And my rant about this recurs for years already. I was not able to verify this as 15.3 has not yet been distributed to my neck of the woods. I'm still at 15.2. In Ps, the detail panel is in ACR - visually different UI from LrC but that's where I expect it. Then I started LrC to see if it had been upgraded already (note to self: makes no sense because it uses the same ACR) and Adobe asked me if I would recommend their software to friends, relations and put a likelihood number on that from the scale 0..10. My answer 0. My answer to the question what they need to do for me to recommend it: "For you to remove the need for me to also buy Topaz DeNoise AI, Sharpen AI, and Gigapixel AI. Or DxO PhotoLab/PureRAW. Adobe Camera Raw's "raw processing" is very bad and I feel cheated on when you release all sorts of parallel products or bells & whistles but don't fix the basics."

@@jpdj2715 My Photoshop and my Adobe Camera Raw plug ins are as far as I can determine up to date and do not have this NR option. So, however it's done, I hope this function comes to PShop/ACR.

@@theprintguide3610 - ACC just updated ACR and I started Ps, then opened a raw file. Ps kicked off ACR and loaded the image. And the Denoise function is there, in the detail menu. 15.2 would open with the detail menu closed, but in 15.3 ACR opens with the Denoise function in the Detail menu exposed.

I have upgraded LrC and tried it out but notice there is a bug - it creates a purple band down the right hand side of the image… image was shot on a Fuji XT2, so xTrans sensor.

@@grantnewton5705 - is that camera property still in the metadata of this image? Is it still "raw"? It was said in several places that this ACR Denoise had Xtrans compatibility, not just Bayer. I cannot verify that claim and do not know if all of the XT cameras are actually supported, but at the moment the statement is "Bayer, Xtrans, raw" and no other formats

Keep it, AI noise reduction is just starting, it will improve in the next few years. But it will need the original RAW file to work

Can I ak if you got any advice? I am running Lightroom on an older PC and a fairly new laptop, and AI Denoise runs very slowly (15 minutes plus per image), so I was going to buy a better specified computer. However, it looks like you have the kind of hardware I would look to get, but you seem to have the same issues,.

@@MrRojo1959 unfortunately, I didn't get any advice. What I've done is changed my workflow. I do all my post editing first and then apply teh AI Denoise last. Actually running the AI Denoise does not take too much time on m system, its just editing the denoised image.

@@FSEGresearch999 thanks. Sorry for the late reply - the notification went to my Gmail account which i hardly ever read!

Yep I get the purple band on my Fuji XT2 image I tried

I compared LrC enhance (Denoise and raw enhance) vs DXO PureRaw 3 for a low ISO image and the PureRaw is far superior, but due to DXO handling the demosaicing much better …. Just has much better detail on xTrans images than LrC manages. Will be interesting to see how high ISO images compare …..

Bayer, not bare. It's a type of sensor.

I max out at about 45 seconds

I think she addressed this in the video. If I recall correctly, she said do a Denoise first, then export to a TIFF then do a Super Res on the TIFF.

Nikon files are .nef, so a .dng would already be a processed file. You have to use the original .nef.

Most current cameras use Bayer petter arrays, most Fuji camera's use X-Trans.

Same here. Nikon D850 file. Large file or incompatible file?

@@photoflickThis is likely your GPU (or lack of a decent one). My computer has a RTX 2070 in it and it does my 45 MP Z9 files in 15 secs. Yup seconds.

@@hchandless Thanks for feedback. I use Intel I7-8700K @3.70GHZ, Navida Quadro and 64GB RAM. But It's still a slow process.

​@@hchandlesssame. 5 seconds with my 3080.

shit sucks takes forever

It only takes 5 seconds for me. I have a very good, but by no means top end computer.

@@joshs7642 Iv got a newer HP Laptop! Lightrooms Ai stuff is stupid fast on it.

@@MrDan1509 weird. I wonder why denoise is so slow. What GPU do you have?

​@@joshs7642 Here's my system info: Processor AMD A6-6310 APU with AMD Radeon R4 Graphics, 1800 Mhz, 4 Core(s), 4 Logical Processor(s)

@@MrDan1509 probably the main issue is your integrated graphics card. I believe the denoise function makes significant use of the GPU. A good dedicated GPU would be much faster