I tried Topaz Denoise for a couple of months, but I was not satisfied because I saw numerous artifacts in small areas. I tried DXO Photolab and was extremely satisfied, both with the quality of the noise reduction and the optical correction because of the huge database of lenses. It allows you to get great JPGs or export a corrected DNG to use in camera raw. I would love for you to try it as your friend Peter did. Have a great day
Wow, thanks for this, Matti. I already use and enjoy the whole Topaz Labs suite so I'm downloading the beta of Photo AI and trying it out right now! Really impressive stuff. :)
Totally agree with your conclusion. I’ve always complained about the low MP of my LX100 II. With the help of AI, I could upscale a few images and keep the rest at smaller file sizes.
really interesting matti, its not something i would need but it sure shows how foto technology just keeps advancing. thx for this video.....zen billings canada
Hi Matti - the "issue" about needing for the programs to transfer the image between each other via TIFF is a common query. Raw is not edited data - it is sensor data plus separate metadata (cooking recipe) for an image. LRc loads the raw data from the file, then you make editing adjustments which have to be stored in the LRc catalog and optionally XMP files. After this LRc has to pass the image to another program. That program has no LRc engine to render the composite image from the cooking recipe so LRc has to pass a cooked TIFF file to the other program. Same idea in reverse. If the other program edits the raw file, it does not save the editing instructions to the raw file, it saves is in some sidecar fille. For LRc to receive the finalised output, the other program has to throw a coooked TIFF file to LRc. LRc has no engine to execute the other program's cooking instructions
I know it's an advert basically, but still, I found DxO's solution a lot better when it comes to detail (compared to Topaz DeNoise). And Topaz turns my RAWs from the G90 darker and red, and scrambles EXIF data. Haven't tried it in months though.
I wrote a comment below, but I don't want it to sound negative. I really like Mr. Sulanto's comments and trust they are honest. That being said, maybe a comparison is in order. I wonder which one would win. However, for me it's moot. I will need to upgrade my computer before doing any AI.
There are still few bugs to work out, but so far, I find Topaz Photo AI to be borderline magical! I've even been able to resurrect even technically mediocre low-res/low-light/high-ISO JPGs from the early 2000s (many of them 1,280 x 960, which is 1.2 MP!). My usual workflow is to start processing with DxO PhotoLab 5 Elite and then export to disk (generally TIFF or JPG) and open in Photo AI. I then see how the AI handles things and maybe then do a little more manually (such as upscaling the "res"--usually Enhance Scale 2 is enough to get a great result and going higher produces some impressively large files). As you point out, Enhance does a great job of interpolating pixels to produce "virtual" high-res images, but it can't create details that don't exist and sometimes smaller faces in the background look a bit strange after AI processing (if they were severely pixelated in the original--there are limits as to what AI can do). That said, there are very few cases where details such as the blind blades in the windows of you example really matter anyway--they won't make the image better artistically or more impactful. This is why it makes no sense to shoot high-res (significantly above around 20MP), unless you have some very specific reasons to do so--for "normal" photography (even landscape, wildlife, and sports), anything much above 20MP is simply overkill. If that weren't the case, then why would Canon and Nikon have "only'" 20MP in their flagship DSLRs when they already produced much-higher-res models?
My computer doesn't have a discrete GPU, and therefore can't take advantage of any AI program. Consequently, I like DXO's non-AI programs the best. I wish I had a better computer, but I live in Peru where name-brand items are twice as expensive and bare-bones to boot.
Thanks as always for interesting comparison, but a question. If I followed correctly, the individual window blinds did not appear in original image from the GX880? No expert--& you certainly are--but I'm surprised the blinds weren't visible even if GX880 is not best MFT camera; as you stated, the Panasonic 20mm f1.7 is a sharp lens. Also what was exposure for MFT photo, including ISO? After watching your video, I'm happy I recently purchased a S5 even if MP is lower than A7 IV.
Näitä lisää....toisaalta olisiko mahdollista vertailla uutta Linux -pohjasta Darkroom 4.0 esim. lLightroomiin. Tiedän, että se olisi työläs tehdä, mutta olen miettimässä siirtymistä Linuxiin ja ainakin tuo Darkroom on aika nopeasti opittavissa vanhalta LIghtroom -käyttäjältä...mitä olen alustavasti sitä testaillut....
Honestly, the improvements you’re showing don’t look great. My guess is that they are not surviving the video downsampling and compression. It’s sort of like those TV commercials that try to show you how sharp or vivid the new TV will look - of course while viewing it on your old TV.
What can you use with Linux? I’m currently using GIMP to save a little. All these subscriptions add up after awhile - especially when a bag of groceries twice a week costs me $80.
I tried Topaz Denoise for a couple of months, but I was not satisfied because I saw numerous artifacts in small areas. I tried DXO Photolab and was extremely satisfied, both with the quality of the noise reduction and the optical correction because of the huge database of lenses. It allows you to get great JPGs or export a corrected DNG to use in camera raw. I would love for you to try it as your friend Peter did. Have a great day
Thanks for sharing.
The job it did in your thumbnail was amazing!!!
Yeah, that is😀
Wow, thanks for this, Matti. I already use and enjoy the whole Topaz Labs suite so I'm downloading the beta of Photo AI and trying it out right now! Really impressive stuff. :)
Glad to help!
Totally agree with your conclusion. I’ve always complained about the low MP of my LX100 II. With the help of AI, I could upscale a few images and keep the rest at smaller file sizes.
Thanks for sharing.
really interesting matti, its not something i would need but it sure shows how foto technology just keeps advancing. thx for this video.....zen billings canada
Thanks.
To use filters with RAW files in Lr, use File > Plug-in Extras > Topaz Photo AI > Process with Topaz Photo AI. It returns another RAW file.
Hi Matti - the "issue" about needing for the programs to transfer the image between each other via TIFF is a common query. Raw is not edited data - it is sensor data plus separate metadata (cooking recipe) for an image. LRc loads the raw data from the file, then you make editing adjustments which have to be stored in the LRc catalog and optionally XMP files. After this LRc has to pass the image to another program. That program has no LRc engine to render the composite image from the cooking recipe so LRc has to pass a cooked TIFF file to the other program. Same idea in reverse. If the other program edits the raw file, it does not save the editing instructions to the raw file, it saves is in some sidecar fille. For LRc to receive the finalised output, the other program has to throw a coooked TIFF file to LRc. LRc has no engine to execute the other program's cooking instructions
Thanks.
in the Lumix photo,
it is possible that the Topaz program sees those fine lines in the windows as moire' and erases them
No, because those fine lines do not exist in the Lumix photo. Topaz can't create those lines from scratch.
I know it's an advert basically, but still, I found DxO's solution a lot better when it comes to detail (compared to Topaz DeNoise). And Topaz turns my RAWs from the G90 darker and red, and scrambles EXIF data. Haven't tried it in months though.
I don’t think DxO has upscaling capability, right?
@@aviatorman8 Haven't got a clue, I'm only interested in denoising.
Yes. DxO is an other planet about noise reduction
I wrote a comment below, but I don't want it to sound negative. I really like Mr. Sulanto's comments and trust they are honest. That being said, maybe a comparison is in order. I wonder which one would win. However, for me it's moot. I will need to upgrade my computer before doing any AI.
Thanks. Your other comment was not negative😀 It's true that you probably need a fairly new computer to effectively use these new apps.
There are still few bugs to work out, but so far, I find Topaz Photo AI to be borderline magical! I've even been able to resurrect even technically mediocre low-res/low-light/high-ISO JPGs from the early 2000s (many of them 1,280 x 960, which is 1.2 MP!). My usual workflow is to start processing with DxO PhotoLab 5 Elite and then export to disk (generally TIFF or JPG) and open in Photo AI. I then see how the AI handles things and maybe then do a little more manually (such as upscaling the "res"--usually Enhance Scale 2 is enough to get a great result and going higher produces some impressively large files). As you point out, Enhance does a great job of interpolating pixels to produce "virtual" high-res images, but it can't create details that don't exist and sometimes smaller faces in the background look a bit strange after AI processing (if they were severely pixelated in the original--there are limits as to what AI can do). That said, there are very few cases where details such as the blind blades in the windows of you example really matter anyway--they won't make the image better artistically or more impactful. This is why it makes no sense to shoot high-res (significantly above around 20MP), unless you have some very specific reasons to do so--for "normal" photography (even landscape, wildlife, and sports), anything much above 20MP is simply overkill. If that weren't the case, then why would Canon and Nikon have "only'" 20MP in their flagship DSLRs when they already produced much-higher-res models?
My computer doesn't have a discrete GPU, and therefore can't take advantage of any AI program. Consequently, I like DXO's non-AI programs the best. I wish I had a better computer, but I live in Peru where name-brand items are twice as expensive and bare-bones to boot.
Thanks as always for interesting comparison, but a question. If I followed correctly, the individual window blinds did not appear in original image from the GX880? No expert--& you certainly are--but I'm surprised the blinds weren't visible even if GX880 is not best MFT camera; as you stated, the Panasonic 20mm f1.7 is a sharp lens. Also what was exposure for MFT photo, including ISO? After watching your video, I'm happy I recently purchased a S5 even if MP is lower than A7 IV.
Those blinds were just too small for that 16Mp resolution. The Lumix exposure was F4, 1/1600s and ISO 200.
It is tempting. But on the other hand, if you need it only occasionally, it's a hefty price.
Yes, agreed. Anything is expensive, if you don't use it😀 However, the price for just the Photo AI is probably less than the bundle.
Wish they would make an iPad app.
Näitä lisää....toisaalta olisiko mahdollista vertailla uutta Linux -pohjasta Darkroom 4.0 esim. lLightroomiin. Tiedän, että se olisi työläs tehdä, mutta olen miettimässä siirtymistä Linuxiin ja ainakin tuo Darkroom on aika nopeasti opittavissa vanhalta LIghtroom -käyttäjältä...mitä olen alustavasti sitä testaillut....
Honestly, the improvements you’re showing don’t look great. My guess is that they are not surviving the video downsampling and compression. It’s sort of like those TV commercials that try to show you how sharp or vivid the new TV will look - of course while viewing it on your old TV.
I use Linux. I have different methods of upscale and denoise. Topaz looks good. It just isn't for me.
I'm sure there are many other methods😀 Thanks for sharing.
What can you use with Linux? I’m currently using GIMP to save a little. All these subscriptions add up after awhile - especially when a bag of groceries twice a week costs me $80.
@@johntravena119 Gimp is good for upscaling. Neat Image is $40 buy if you are a pro, free if you are a non-pro (unpaid) like me for de-noise.
@@tonykeltsflorida Thanks, I’ll check it out!