Compare denoise in Lrc, Topaz and Photoshop

Status
Not open for further replies.
A

adonetok

Member
Joined
Aug 29, 2020
Messages
89
Operating System
  1. Windows 11
Compare denoise in Lrc, Topaz and Photoshop, which one is the best?

I use denoise in Topaz (From Lrc) before.

I am learning Ps now.

Is denoise in Ps really better than Topaz?
 
Sort by date Sort by votes
I was not sure if I should post this as there are Topaz users here. I did look her up and she's not some basement hack. I don't think so but of I'm wrong let me know. Just to point out Topaz is a good product and no two files are alike.

I have come across other videos that also show Topaz sometimes doing weird things in specific areas of a file. I can't compare to what is current. I only have Photo AI v1 but I did see this happen with that version. This is 10 months old, Topaz is always improving so she may have a more current video.

https://www.youtube.com/watch?v=DbJLp-AxHWM
 
Upvote 0 Downvote
KarstenG said:
How long approx does one photo need to denoise in LrC? And what's the time required for this by Denoise?
I'm asking because I have recently upgraded my GPU and found that all denoise tools (LrC, DxO and Topaz) need similar time - which depends on GPU power,
Click to expand...
Has no one has reported 8 sec Denoise times in LrC. This is what I am seeing after I upgraded my 16GB M1 iMac to a 64Gb M2 Ultra Mac Studio. I found the 30s times I was getting with the iMac to be unacceptable.
 
Upvote 0 Downvote
I can’t remember if I posted this regarding Adobe Denoise AI. My 2019 Intel takes 33 seconds. My 2020 MacBook Air M1 with 16GB ram used to take 85 seconds. After Apple fixed the Neural engine problem it dropped to 25 seconds. That was at the 13.2 release, I think.
 
Upvote 0 Downvote
clee01l said:
Has no one has reported 8 sec Denoise times in LrC
Click to expand...
For me it's even less - for my 20 MP OM-1 RAW files :)

I have asked the TO how long it takes because he stated "I try Denoise in Lrc but it takes too long to complete it."

I my experience the processing time is very similar for LR Denoise, DxO Deep Prime or Topaz Denoise/Photo AI.
Before I upgraded my GPU to RTX 4070 it was ~45 sec for any of them, after the upgrade it's 5s.

So, if Lr Denoise is taking too long for the TO, it should be same or at least very similar for the other apps. Provided same performance settings.

@adonetok :
What CPU/GPU are you using?
 
Upvote 0 Downvote
On my iMac Intel. Topaz Photo AI v1 took 45 seconds. They may have improved performance as it is v3 now. DXO DeepPrime took 25 seconds. DeepPrime XD took 45 seconds. Those were about the same times on my MacBook Air M1 but I'm not sure about recent performance issues. I know DXO also issues with the Silicone chips but it was colour. DXO had a work around pretty quickly but colour is a lot easier to fix.
 
Upvote 0 Downvote
@Gnits It would be interesting to hear of your frontline experience.

Sorry, but going further off topic I know. I've encountered people from the 'always Ps' camp some of whom even ran intros into Ps but I found they were teaching the same approach using the same tools the way they had been doing so for many years although they were using the latest version their Creative Cloud subscription offered. When asked about new tool x or y the response was a very blunt dismissal, I've been doing it this way for years and I've no time or interest in doing things differently. I find they are the same ones who think M is the only setting on a camera that should ever be used.
 
Upvote 0 Downvote
iwaddo said:
It would be interesting to hear of your frontline experience.
Click to expand...

I'm not Gnits but I'm using AI Denoise by Adobe, Topaz and DxO to some extent in parallel, so I guess I can provide some frontline experience too ;)

I started using KI based Denoise SW in 2020 (Topaz Denoise AI). I was absolutely amazed how this SW could improve the technical quality of wildlife (birds) images taken with my crop-2 sensor camera. DxO PhotoLab even surpassed the level of quality, especially for images at higher ISO and those with relevant details in darker areas.

When Adobe released AI Denoise I was amazed again, because I did not expect their solution to be on par with Topaz and DxO, but in my experience it is.

Today, AI Denoise in LrC is my #1 approach over Topaz and DxO PL because it delivers high quality results which fully satisfy my expectations in most cases. For this, I typically do not even change the "amount" parameter (40 is my go-to value). It's a one-touch processing in my RAW development tool which is done in a few seconds.

With both Topaz (now Photo AI) and DxO PL7 I can accomplish very similar results, but this usually takes 1-3 iterations because both apps have a tendency to oversharpen using default settings. Even with optimal settings for Topaz and DxO, to my taste LrC Denoise renders find details (fur and feathers) better than Topaz and DxO, in a very natural way.

Effectively, I'm using LrC Denoise for the majority of my images. If I'm not happy with the result, I am doing another round in DxO PL. This often involves images that also lack sharpness. In these cases, PL's oversharpening is welcome.

My only use cases for running Topaz Photo AI are
a) trying to rescue technically failed captures (by fixing motion blur or out of focus problems),
b) resizing massive crops
c) providing a crisp appearance at 1:1 level to non-RAW images like high resolution interior panos to support a "virtual tour" experience

All this is based on my experience with the images resulting from my images (which can be seen on my Flickr) account.
 
Upvote 0 Downvote
I’ve been watching a lot of videos and there are reviewers that say Adobe has a more natural look. I’ve even seen that on photo forums.

After Denoise came out I dumped all my 3rd party NR apps but kept I kept Topaz Sharpen AI for the same reasons as you. I only use it if I think the file will benefit. I don’t keep too many OOF files. If I do it has to be pretty special. I really hope Adobe puts out a Sharpen AI module some day.
 
Upvote 0 Downvote
Forgot to say. Some of the reviewers stated they liked to start at 43 so that’s what I use.
 
Upvote 0 Downvote
KarstenG said:
Why do even several people recommend such a crooked number as a starting point for AI Denoise?
Click to expand...
Empirical findings? Trial and error? Round numbers appeal to us humans as a nice simplification of a mental model that we might be holding. But the real world often does not line up with the number system used by ten-fingered and ten-toed bipeds. :)
 
Upvote 0 Downvote
KarstenG said:
43 !?? I know that "42" is a special number ;)
Why do even several people recommend such a crooked number as a starting point for AI Denoise?
Click to expand...
I don't really know. Testing, etc as pointed out. 50 might be getting into the plastic look. It's my start point which doesn't mean it's where I wind up.
 
Upvote 0 Downvote
Zenon said:
I don't really know. Testing, etc as pointed out. 50 might be getting into the plastic look. It's my start point which doesn't mean it's where I wind up.
Click to expand...
I almost always use 80 and generally I am pleased with the results. I think it might depend on how much color noise is present and perhaps what ISO setting was used to take the photos as high ISO numbers tend toward noisier images.
 
Upvote 0 Downvote
If only to add to the dialog. Has anyone an opinion on the new TopazAI "Super focus"? Which I find incredibly slow on my M2 Ultra MacStudio but incredibly good.
 
Upvote 0 Downvote
clee01l said:
I almost always use 80 and generally I am pleased with the results. I think it might depend on how much color noise is present and perhaps what ISO setting was used to take the photos as high ISO numbers tend toward noisier images.
Click to expand...
I'm always trying to squeeze out as much feather detail as possible. I can't afford a big f4 prime so I have to work with a slower lens.
 
Upvote 0 Downvote
Status
Not open for further replies.
Back
Top