How to eliminate noise with PureRAW 3

[@NickPage]

Really impressive noise reduction to be sure, I don't care for the look of the sharpening algorithm, especially so early in a workflow, but that noise reduction is intriguing.

[@gregbenzphotography]

Agreed, would be nice to have full control to isolate the noise reduction from the detail enhancement in cases where it isn't ideal.

[@MartinHarvey]

Thanks for the excellent prez and good to see, Goosenecks again! I have resisted the impulse to try and cough up the money to acquire PR in the past as I do have DxO and the NIK filters (which I've had from its long past get-go). I primarily use Adobe LR Classic as my mainstay and the peerless LR Mobile when I travel. I have used DxO as a Pure Raw alternative as it shares the same denoise controls and I often complement this with Clearview before returning to LR Classic. I also have Topaz Photo AI. However, PR has a different role it can play early in the editing workflow and subtle differences from using DxO only. My next big trip is in March so maybe I'll try and PR free trial when I return. I do not print images but do create large format Blurb books via the LRC book module for each trip we take. FYI my equipment is an X-Pro2, GRIII, and an iPhone 13 mini. DxO doesn't support phones BTW!

[@irampotter]

I used this on a shot of a boat race I took the other day, in the original file you could make out the individual people in the background crowd but in the processed file you can really distinguise between the people, the shot was a 50 MP file shot at 1600 iso 1000/sec at f.8, I was happy with the original until I read your email and tried this software out, the difference is really significant, thankyou.

[@gregbenzphotography]

That’s great! Thank you for sharing your experience with it.

[@J5388T]

Greg, Thanks for posting what is a very informative yet unbiased review, you're the first person to do so on YT whose opinion I respect. I could see the difference in your comparison shots on YT even before you zoomed in and as you say they're impressive. I currently use PR2 as my preferred noise reduction software and whilst I'll download the trial initially I can see me upgrading to PR 3!

[@gregbenzphotography]

Thanks, John!

[@KevinMullett]

The day DxO PureRAW 3 came out, nobody had anything on aurora or Milky Way shots so I downloaded the demo. Glad to see someone talking about the star situation as that lattice effect it is creating is a deal breaker (as is) for large prints. I will probably dual process as you mentioned. At internet sizes it is mostly imperceptible. I also had a strange warping effect happen on a shot with a curved 6x6 timber staircase that lead to the Milky Way, but when it works, wow, does it work.

[@gregbenzphotography]

I’ve also seen that the distortion correction can affect output. One image showed an artifact that looked just like a shooting star when I used the max rectangle method. Maybe the AI saw something in the stretch that triggered a defense to a shooting star. Or maybe it was just a straight white line for some reason. Only one like that though, not common.

[@KevinMullett]

@Greg Benz Photography maybe I just have not hit on the right combination of settings that doesn't trigger that patchwork lattice issue.

[@gregbenzphotography]

To be clear, I definitely have star issues in both v2 and v3 (with v3 being worse and v2 suffering more for making secondary stars too prominent). I just have also seen a faking shooting star in one case using v3 with distortion correction (it’s repeatable but just with that one image).

[@wegder]

I downloaded a trial and it is amazing so far, I wish I understood what is going on under the hood.

[@KaReEdCa]

I am impressed with Pure RAW3. it clearly reduces noise way beyond anything I've seen before. I do have an issue when people are in the shot. It tends to oversharpen them, leaving visible lines (almost like a video game). I need to experiment with the plethora of options in this new version to see if I can improve the outcome. For landscapes it's a no brainer!

[@gregbenzphotography]

I’ve had very good results with a series of ISO 800-1600 event shots, but it will depend on how much the person fills the frame, etc. Have you had a wide range of people you’ve tried and seen issues, or just a few so far?

[@KaReEdCa]

@Greg Benz Photography I've not tested extensively... one an event shot and another. a headshot.

[@tamashaasz]

Thanks for posting this Greg! So far it is the most informative review of PR3. My test running PR3 vs PR2 vs Topaz Denoise (latest standalone app, not the Photo AI) with Nikon D850, same NEF file, shot at ISO 10000, shows that PR3 has less noise and more detail than PR2. Comparing the Topaz Denoise results with PR3 and PR2: both produced more artefacts than Topaz on the same NEF file. My second test on a different set of images (D850, NEF, ISO 4000) yielded better results with PR3 and PR2 vs Topaz. Similarly to the example in Greg's video, visible gain on details preserved with no noise present-> PR3 vs PR2. This time Topaz produced artefacts in some areas of the images, whereas, no such issues with PR3 or 2. As for me, PR3 is the first tool I would use to complete noise reduction. It makes PR2 redundant since it can run the older AI models (Prime and Deep Prime) and the new Deep Prime XD too. Now it has the control over lens correction, softness, CA etc. Last but not least, the integration with Lightroom (and OSX Finder) is very user-friendly, plus it keeps all adjustments previously made on the RAW file. In case of errors (such as artefacts), my next step would be to try Topaz Denoise and see whether it's producing better results.

[@gregbenzphotography]

Thank you for sharing your test results / experience

[@YohannChaplin]

Thank you Greg, I was wondering if I should get PureRaw 3 and you definitely convinced me! They have a bundle with Nik Collection 5, but I wonder if they will release Nik Collection 6 in June?

[@gregbenzphotography]

No idea on future releases (and wouldn’t be able to say if I did)

[@paulmoore721]

Fantastic review! I have the software and was wondering how it would work with low-ISO images.

[@gregbenzphotography]

Thank you! Yeah, that was a really nice surprise to me to see noise reduction so good that it's highly compelling at low ISO.

[@jiml989]

I have version 2 and am currently comparing with version 3. I'm finding some images are margianlly better but also in some images there is significant artifacts appearing. I understand that by turning on and off options, comparing different results, etc., could result in the "ideal" settings for a specific image. Frankly, however, I don't see the marginal improvement is worth the work flow cost. I'll continue to experiment, but so far it's not what I had hoped for.

[@gregbenzphotography]

v3 is definitely a little more prone to artifacts (though you can run the v2 algorithm from it, so you're never losing that option). It tends to add more detail than artifact, but there is some degree of tradeoff. Are you not seeing much improvement with noise? I've seen huge gains on images at any ISO and exposure level I've tried. I think the noise reduction is the most significant improvement here.

[@jiml989]

Thanks so much for your reply. I guess I'm in the minority but what I see, even on some of your examples (especially foliage) seems more mushy than I would prefer. I prefer V2. My skys are definetly less noisy in V3 but I can use masks in ACR to apply more denoising if/where I need it. Again, for me, it's just that the squeeze isn't worth the juice. FWIW, I'll be the first to say you are much more experienced and undoubtedly see things at a different level than I.

[@gregbenzphotography]

Try playing with adding some noise. I think you’ll find that the nose can be visually mistaken for detail (it’s not always bad to have noise).

[@jiml989]

Interesting, I'll keep plugging away at it. Not giving up yet. Thanks for this and all the great content you create.

[@Daniel_Zalman]

To me, the newer version sacrifices some detail for the increased reduction in noise. If you at 1:55, there's more edge detail on the image on the left. Then again, the second is noticeably better in version 3.

[@gregbenzphotography]

I see what you're saying there. However, that's not been my experience with it. Noise can create the appearance of fake detail, and I believe that's a factor in the comparison at 1:55. I went back to review the foreground of that aurora image just now. If I turn on some modest sharpening for the v3 image, it shows similar levels of detail while staying at vastly lower levels of noise. Or if I turn on the various v3 sharpening options I also see more detail (however, the tree line starts to show minor halos and I would blend different v3 outputs if I wanted to use this approach for more foreground detail). In either scenario, I feel the final print (or even social post due to noise) would look better coming from v3. That said, this image is generating a lot of content in a severely limited high ISO shadow area. Most images have more real content and I've found that v3 holds up even better in those cases. The only case where I've found v3 isn't as good as v2 is in stars. And in that case, there are some other approaches I prefer to both versions as v3 shows unwanted artifacts / too many minor stars and v2 doesn't remove enough noise (and shows some similar problems as v3 but to a lesser degree).

[@Daniel_Zalman]

@@gregbenzphotography Thanks very much for the detailed response.

[@gregbenzphotography]

Absolutely. And thank you for your question, I always appreciate discussions that make me to think a bit.

[@JDon8110]

way cleaner obviously, but looks like loosing a heap of detail too between V2 and V3. Hav you done a comparison between this and topaz

[@gregbenzphotography]

I don’t really believe that’s the case from what I see. Probably a combination of video softness (not native screen resolution) and noise appearing like detail. You can easily sharpen or add noise. Not extensively, but my impression is that PureRAW does a much better job. I’d recommend giving DXO’s free trial a go to see what you think.

[@benjaminnorton9561]

Have you ever tried the topaz labs noise reduction against the DxO? Been using topaz for a couple of years now but I’m starting to think there may be a better option out there.

[@gregbenzphotography]

Yes, and I feel PureRAW 3 is better. DXO has a free trial. gregbenzphotography.com/photography-reviews/eliminate-noise-with-dxo-pureraw-3-deepprime-xd/

[@carlkristensen8398]

Have You tried to compare it with the Topaz Photo AI?. I very much like what I see here, but I allready have the Topaz, wich I think is great running as PS plug-in. It is just a never ending purchasing business, on the oter hand one would try to get the best possible "outfit". It looks a little like they are doing the same - but I can see the advantage of keeping it in the original raw file. What are your toughts?

[@gregbenzphotography]

I feel DXO does a much better job. Take that Aurora image I showed in the video for example. After processing with Topaz, I still get a DNG (which I greatly appreciate). However, it doesn't work normally. The response to sliders is totally different. I cannot lighten the shadows in any realistic way, they are nearly black. And when I lift them through exposure, the shadows show a very heavy purple cast and effectively unusable. There also is not as much detail. I've seen other cases where it does a nice job. But on the whole, I trust PureRAW to produce something which is nearly always at least as good and often better. But every type of image can show a different result, and I would recommend you use their free 30 day trial to try for yourself on your own images.

[@carlkristensen8398]

@@gregbenzphotography Yes, I might try that 😀

[@bmichaud1000]

Thanks Greg. Have you compared the noise reduction results from PureRAW 3 to an image whose noise has been reduced through stacking multiple images?

[@gregbenzphotography]

For night sky images, I would stack. The result is clean and avoids artifacts. That’s the one area where I’ve not had success with PureRAW 3. In an ideal world, I would use PR3 before stacking, but the artifacts and enhanced minor stars I see are too much. It makes for a cluttered sky. I expect that day will come when we can combine them (either with a future version of PureRAW or some other tool targeting a similar RAW workflow). It would be really nice to do something like a smaller stack of 4 images with RAW noise reduction that trying to shoot and stack 10+ images.

[@bmichaud1000]

@@gregbenzphotography Thanks for the advice. I'm heading out tonight to take advantage of the new moon. I may pick up the software anyways to play around with it. Much appreciated!

[@boruchlen]

Very good comparison video, thank you! If I may ask, why is there a white fringe on the edge of the sky/mountain at around the 4:05 mark? I get that often, and I was wondering how to get rid of it. Thanks again for all your great videos and tutorials.

[@gregbenzphotography]

DXO is adding some sharpening at all times (at least as of v3.1). You can blend with another version of the image, potentially using Adobe's new AI Denoise: gregbenzphotography.com/photography-reviews/ai-denoise-lr-acr/ You can use the "Edge" button in Lumenzia to help paint masks or similarly as a selection to help guide cloning when you need precise control over such a fix: gregbenzphotography.com/luminosity-masking/how-to-remove-exposure-blending-halos/ I use a mix of these sorts of techniques with a filter or AI causes problems around edges, which is fairly common.

[@boruchlen]

@@gregbenzphotography thanks so much for your detailed response, I'll check out both links. Keep up the great work.

[@gregbenzphotography]

Happy to help!

[@Rocking_J_Studio]

I currently use version 2 and I use it on all of my images. However, I shoot an average of 25-30 images per composition when focus stacking which I do on a regular basis. I use Helicon Focus to merge the images into the focus stacked version. Where in the workflow would you recommend using PureRaw when focus stacking images?

[@gregbenzphotography]

You can only use PureRAW on a RAW file which has not yet been demosaiced (for example, you have to run it before using LR's merge to HDR/pano, but can use it before that step). So you will need to use it before Helicon Focus. You can apply slider values to the RAW before Helicon and I would (unless they've added some support for exporting a merged RAW that I've missed).

[@joelbarto1103]

Greg, Can you apply the DXO 3 after you've processed? Say after you've blended a 2 minute foreground with a stacked sky image? Are there benefits for using it before or after?

[@gregbenzphotography]

It requires a RAW. You can use a file with adjustments in LR, but definitely not something like a TIF image. No real benefit to using before or after making adjustments in LR, they get ported over nicely and respond the same way (not true of all RAW enhancement tools and there may be some edge case out there I'm overlooking - the only problem I've seen was with a local correction for color which needed some tweaking after conversion: gregbenzphotography.com/photography-tips/recover-noisy-shadow-detail-with-dxo-deepprime/)

[@KGi4]

Good to know the power of such a tool but, does it perform better then Topaz deNoise Al version?🤔

[@gregbenzphotography]

See my other comments, I feel it does

[@jensvielmann7662]

I trust you, so I assume that is because of KZbin and may be my display… But honestly that didn’t look all that impressive to me. It looked like it may be a tiny bit better denoise algorithm. Especially the image with the trees. They look very soft and I would rather say lost a bit of details than gain some. plus some of the tree clusters that you didn’t zoom in to look like one big bloop of color. So basically the same problem you have with a D noise you’ve got to sharpen it afterwards. And when you do, you’re probably not too far off from the original. Maybe a tiny bit in the rocks from the second image, but even that look like it needed some sharpening.

[@gregbenzphotography]

The differences are quite significant on my screen. So much so that the aurora image would show benefits for some social media exports for noise reduction and all would benefit in a print. I recommend you try their demo for yourself. Resizing of pixels for KZbin makes the video not give you the same clear different I see on this end (probably more similar to my screen in segments where I resized >100%). I do not see any appreciable loss of detail when using the tool with proper settings and potentially sharpening after if needed. Noise can look like detail, so it can throw you a bit, and you can always add noise if that's an effect you want. But the gains here are substantial in my opinion, even over v2, and certainly over the original RAW. The tree detail in that aurora image is extremely dark. The "bloop" is better than the mess of noise in the original. the original image is even more lacking in detail, it's just harder to tell with the noise masking it. So that might be an area where you'd allow a bit more noise or blend results if you want to obscure imperfections in the file with noise.

[@ptatore]

How does this software compare to Lightroom's enhanced Denoise.

[@gregbenzphotography]

See gregbenzphotography.com/photography-reviews/ai-denoise-lr-acr/

[@richarddeyoung8129]

My Canon R5 lets me set auto lens correction in camera for both chromatic aberration and profile corrections. Is it better to use Canon's own in camera corrections, which show up in the Lens Correction panel in Lightroom/CR or the DxO post corrections? If I apply the in camera correction and then the Dxo RAW correction on top of that, does that create problems or a less correct image?

[@BransBowler]

I’m not certain but I think that those in camera correction settings are only applied to jpegs and on camera previews.

[@richarddeyoung8129]

@@BransBowler No, when I process the RAW file in ACR or Lightroom, the boxes are prechecked for the adjustments. Not sure it Canon is actually making adjustments or just telling ACR to check those boxes. Thanks for the reply.

[@gregbenzphotography]

I'm not familiar with what they are doing. If they are applying their own corrections on the RAW, I would expect it may be better given their proprietary knowledge and focus. But if it is not RAW with corrections, it's probably not better. Additionally, PureRAW requires the original RAW and may not work on any such RAW which was processed (just test with their free demo to see if it works if you need to evaluate before purchase). Personally, I prefer to correct in LR so that I have the option to blend with the original RAW or control the degree of correction (I often use some middle value rather than full correction to get the look I want).

[@richarddeyoung8129]

@@gregbenzphotography I have processed RAW images with the Canon applied settings for the lens with DxO Camera RAW settings also applied and it seems to work; I am need to test a bit more to see if it overcorrects that way. I have also turned off the corrections Canon has applied in ACR before processing the raw file in DxO and that seems to work too. Just FYI, when Canon sets the corrections in camera, they are still adjustable with ACR sliders.

[@gregbenzphotography]

If you find it over corrects when captured that way, please email me with a sample image with and without that camera setting and I can forward to their engineering team through a contact of mine. They seem very eager to keep improving.

[@hugoparente1908]

Do you think that is better than Topaz Ai, in witch way? I feel that they are running behind...

[@gregbenzphotography]

I know many people love Topaz, but I've found I like what I get out of PureRAW better. I haven't used Topaz exhaustively by any stretch, which is why I didn't comment on it much.

[@richarddeyoung8129]

@@gregbenzphotography I just tested Topaz Photo AI and DxO PureRAW on a photo of a bison cow and her calf. Absolutely no comparison, DxO PureRAW blew away the on processed in Topaz Photo AI as a raw file creating DNG. Topaz still had noise and some artifacts, DxO PureRAW DNG was very detailed and clean. I have used Topaz a lot and couldn't believe how much better DxO PureRAW is.

[@gregbenzphotography]

@@richarddeyoung8129 Thanks for the feedback! That's consistent with my testing so far.

[@hugoparente1908]

@@richarddeyoung8129 I've already tested it and it's really useful when we need to raise the ISO or work on raw files from small sensors. However, it is different from Topaz photo AI, which can be used at the end of processing, unlike Pure Raw 3. The DNGs of my pixel 7 pro, do not run in Pure Raw 3. In short, the two can be complementary in an editing process.

[@rolandrickphotography]

That at 2:01, the DR3 version looks (at least on YT) totally washed out, blurry, unusable.

[@gregbenzphotography]

It's much better, and you can add in noise as needed. Touch to show since you can't see at my native resolution with the online video. The v2 version just has a lot of noise that looks like detail. Easy to recreate that in v3 if you ever needed it (the exact same method is available, but you can do much better with v3 and various lens sharpening methods and possibly adding sharpeness via ACR/LR).

[@MaximDupliy]

Hello Greg. On munute 5 and 25 sec you claim that DXO prime engine is much better at noise removal than ACR, but you don't show the ACR cleaned of noise image to compare. At least show it to us. It's not a good idea to take image with noise without doing noise removal and compare to one that is after noise removal. SHow us how the ACR cleans it and then compare, it will be much more reasonalble.

[@gregbenzphotography]

That's a good suggestion and if I redid the video I'd show that. I'd recommend using their free trial if you want to compare for yourself. I'll add that you're not seeing the video at the native resolution of my monitor, so you're not always seeing the full difference (my ISO64 comparison shows even more benefit on my monitor than the video).

[@howardthompson7661]

how does it compare to Topaz products

[@gregbenzphotography]

Addressed in some other comments here, I prefer PureRAW

[@OnMyOm]

Buyer beware: it doesn’t work with Apple ProRAW files from iPhone. Just a heads up!

[@gregbenzphotography]

Apple ProRAW has already been processed, but the standard RAW has mosaiced data. You'd have to use a 3rd party camera app to capture in that format, as I don't believe the native camera app allows a choice of RAW formats.

[@OnMyOm]

@@gregbenzphotography thanks Greg. I saw you explained that in your latest video. Appreciate it

[@Dr86Jones]

This noise reduction softwares are expenssive and for me they dont worth the money. With the most recent cameras i dont know why do we need this anymore. I think people should stop pixel peeking and focus more in taking good photos.

[@gregbenzphotography]

It’s certainly one of the more expensive pieces of photography software I use, but it also adds substantial value for my final results for high ISO images or for making large prints at any ISO. That’s even when using ideal technique on some of the best 35mm cameras ever made (you can probably outdo this with medium format, but that costs much more than this software). Whether this benefits you or not depends on your needs/goals, which is why I always try to be clear where I think a tool or technique adds value.

[@KaosK9]

I have purchased and used them all. PR3 is by far the best performance and value of all the NR options currently available. Doesn’t matter what the quality of your camera is, there will always be a time where a high quality NR option is welcome. While I don’t use it on most images, I have used it on high ISO images from top grade cameras with amazing results. Thank you Greg for the honest review!