Nikonscan 4.0.3 C41 inversions and RA4 paper colour signature question

[radialMelt]

I'm curious if I can spot the Nikonscan output... 1.2, 2.3, 3.3

 

[albireo]

I'm curious if I can spot the Nikonscan output... 1.2, 2.3, 3.3

I will reveal the labels in a week or so.. Just to give time to anyone else who might want to give it a go. Could be interesting!

 

[radialMelt]

I will reveal the labels in a week or so.. Just to give time to anyone else who might want to give it a go. Could be interesting!

Care to reveal the answers?

 

[albireo]

Care to reveal the answers?

Oh yes apologies. I had forgotten about this. And I was hoping to see some guesses+rankings by people. So disappointed

But first allow me to add one last example. I just realised I had only posted 35mm samples. Here's one from a 120 scan.

Ektar 100 (expired 2019 - cold stored) in a Texas Leica. Same four options as described in post #25. Any guesses? Any personal rankings?


I will add a hint, at this point: I found one of the tested methods really tough to use and almost impossible to squeeze anything good out of. I was surprised actually.


1.

2.

3.

4.

 

[Steven Lee]

@albireo 2nd from the top would be my preference. It looks perfect. The 1st has too much cyan, the 3rd is over the top magenta, and the last one is a good starting point for further editing: the colors are more neutral and it is flatter.

By the way, I find this image oddly appealing. I don't know whether this was a test shot or an intentional composition, but it's an eye candy and strangely relaxing.

This also brings up the question of how to evaluate the performance of a scanning workflow (hardware+software). The best looking and pleasing image may not be the most accurate one. Consider a hypothetical example when a color inversion software "A" makes everything warmer and contrastier than software "B". But this only means that for some scenes "A" would work better, and for others you may prefer "B". I'd probably choose a tool which gives me the least opinionated and most neutral starting point.

I guess I'm saying that post-scanning Photoshop tweaks are unavoidable. In fact, I would argue that it's a good thing, because it explicitly separates the scanning step from the color-grading step. For example, if your preference lies on the warmer/contrastier side, you can simply apply this bias to the entire folder of neutral scans, i.e. you can easily automate it. But if your scanning workflow always gives you a biased starting point, this makes it harder to deviate from that look: both manually and via automation.

So... going back to your images, I would prefer the default scan to look like #4 so I'll have the option to edit it to look like #2

 

[radialMelt]

Of that 120 set, I think #1 (or maybe #4, this is a tougher test) is Nikonscan. Impossible to really tell the others, in my opinion, as it's more color choice than anything. My 2c

 

[albireo]

Thanks @Steven Lee for the comment - very interesting ..

One more example before I spill the beans.

6x6 negative from one of my TLRs. Tough film to invert (IME): Fuji Pro 400H.

1)

2)

3)

4)

 

[Steven Lee]

@albireo I would go for #1 in your last set because of the same reason I stated above: it is the most neutral and it's easy to add just a bit more warmth. The other three have too much "flavor" injected already.

 

[radialMelt]

I'm going guess #3 in that last set is Nikonscan. I only say this because of the awkwardly high white point, which I have always struggled with with Nikonscan, despite its merits.

 

[albireo]

I'm going guess #3 in that last set is Nikonscan. I only say this because of the awkwardly high white point, which I have always struggled with with Nikonscan, despite its merits.

I have to say you're extremely consistent in your predictions so far, but you're consistently wrong - meaning you consistently pick the same method, which is not Nikonscan.

 

[radialMelt]

I have to say you're extremely consistent in your predictions so far, but you're consistently wrong - meaning you consistently pick the same method, which is not Nikonscan.

Haha excellent! Looking forward to the big reveal

 

[albireo]

So the following are the labels for the images shown.

Post #25 - Image 1 - Kodak Colorplus 200
1.1 Vuescan
1.2 Filmomat
1.3 Colorperfect
1.4 Nikonscan

Post #25 - Image 2 - Kodak Gold 200
2.1 Vuescan
2.2 Nikonscan
2.3 Filmomat
2.4 Colorperfect

Post #25 - Image 3 - Fujicolor C200 made in Japan
3.1 vuescan
3.2 nikonscan
3.3 filmomat
3.4 colorperfect

Post #29 - Expired Ektar 100 in 120
4.1 filmomat
4.2 colorperfect
4.3 nikonscan
4.4 vuescan

Post #32 - Fuji Pro 400H in 120
5.1 vuescan
5.2 colorperf
5.3 filmomat
5.4 nikonscan

Here is proof that I own the tools and have actually used them.

Setup
For Nikonscan, I've set the tool to calibrated RBG, chosen the focus point, and produced a 48 TIFF output.
For Vuescan direct, I've left all defaults, 48bit rbg color negative mode, auto white balance, and produced a 48bit TIFF output.
For Colorperfect I've used Vuescan to produce a 48bit positive raw file in TIFF directly in PS.
For Filmomat, I've loaded Vuescan's raw file and produced an output TIFF.

I've then opened each image in Photoshop CS6. I've set myself a constraint via a timer on my phone. I wanted to make sure all of the final jpg images to be shared were ready in max 1 min Photoshop activity starting from loading the raw (or tiff in the case of Nikonscan and Filmomat) to final small jpeg generation.

I did not apply custom color correction, or other edits apart from: crop, set black point to histogram tail, resize, save to jpg.

A kind reminder from post #25:

Note for the tinkerers: I know you can improve any of these with a wise combo of PS color edits and/ord autocolor and/or curves. This is not the point of my exercise. The point of this is to see which method is giving me the most pleasing results out of the box or gets me closest to my memory of the scene, or gives me the output that is simpler to tweak into a final image.

Other observations

• For filmomat, I've followed the tutorial here https://www.filmomat.eu/
• For Colorperfect, after loading the image, I selected the film stocks' colour profile from the drop down box. Note: I was not able to find a matching color profile for the Fujicolor C200: I decided to go for 'Fujicolor 100' out of lack of better options.

Some thoughts

Filmomat Smartconvert
Let's get this out of the way. Based on my hardware, preference, and sample images, I found Filmomat to be the absolute worst of the lot. The images looked all blown out, with weird colours, strong saturation, lack of detail etc upon loading. I did not locate a histogram control to restore the highlights. The colours were, I thought, weirdly saturated, similarly to some of those instagram digital filters people apply on their phone images. Filmomat was the only tool were I did have to tinker a little with the three main controls in the interface: contrast, density and saturation. The images I shared were all saved after choosing increased values for contrast and density and vastly decreased value for saturation.

It didn't really help. None of my results remind me AT ALL of the scenes I saw and I preserve in my memory, and I don't want to spend time at the computer attempting to recreate them out of chaos, so Filmomat is out for good for me.

Colorperfect
This is interesting. Having used it mostly for Kodak film in the past (Ektar 100, Gold 200 mainly over the past few years), I had always been pretty satisfied with it. I was convinced it is a pretty reliably good tool, offering no real surprises. It was good then that I tested it on Fuji material, because seems like I've been proven wrong: the Fuji results are, IMO, pretty bad. Image 3.4 (post 25) has a strong green tint I find unpleasant, whereas the last image (boat, post #32) is plain wrong. I really like, however, what Colorperfect does with the Kodak material, including the expired Ektar roll. Also, it's worth noting neither 400H nor C200 Made in Japan are available anymore (at least to me) so perhaps not a big issue.

I still like it for the great amount of control and the very conservative approach to preserving the entirety of the histogram: notice how its images have more muted highlights and appear perhaps a little 'darker' than others. I still think it could be a good tool for many jobs.

As for the Fuji failures, I wonder if it's down to age: perhaps those film-specific profiles it uses are obsolete? Perhaps current film has undergone variation from whatever data was used by Colorperfect's author to encode all those film.

In general, this little exercise teaches me that its output is not tinker-free. So not one of my favourite options anymore.

Vuescan
I am really impressed by the nice, vivid, but not over the top results given by the vanilla Vuescan inversions. I do detect, for my liking, and based also on other examples, a slight tendency to veer towards the cyan/green on some images (somewhat visible in background of the 400H example, last image), and the 'too lively' tone of some images. But this is something even I would be prepared to deal with with a minimum of post-processing in photoshop if Vuescan were to be the only option available to me. I do find it adds a little too much saturation than I'd like in places. But overall I'm pretty pleased really.

I think based on the consistency shown I'd place Vuescan in a solid 2nd place.

Nikonscan
I see few surprises here and I remain happy about what I can get out of it without any sort of intervention whatsoever. Couple of observations

• my memory that Nikonscan did a really poor work with the highlights was not entirely justified: I find the highlights in the images shown overall pretty pleasant, thought I would love a little more control on them. But the effect is not as bad as I remember. Filmomat does much, much worse with the highlights
• With the expired Ektar 100 sample, I notice a red/magenta dominant, and I've noticed Nikonscan producing this often with expired Ektar. I have not noticed this with other rolls of fresh Ektar developed recently. So I think the Ektar sample is a bit unfair to Nikonscan. This is a rare case in which I'd bother tweaking the magenta slider a bit in PS, because I find the rest of the tonal balance for the autumnal Ektar scene pretty spot on and very close to my memory of the scene. However this is one of those examples where Colorperfect does perhaps a better job out of the box.

So I hope this was interesting for someone. Also - I realise I had promised to compare NLP too, but I will do so in the future hopefully: I'm on my own (old) PC in my home country and I don't own a copy of Lightroom here.

 

[radialMelt]

Thanks for this! My preferred tool as of late has been Filmomat SmartConvert (w/ vuescan raws) due to being quick for adjustments. I really like how it the UI is based on lab scanners. The lack of historgram is pretty unfortunate. I've mentioned this to the developer. Hopefully they add it in a future release. It is very very ironic that I picked it every time out of the bunch, as I was looking for the worst conversion and calling that Nikonscan haha!

You've inspired me to give Nikonscan another go. I have to admit I struggle immensely with consistency no matter what tool I use, be it NLP, Vuescan, Nikonscan, or SmartConvert. Seems like some frames convert really nicely, and some are awful, even when the scene is the same. I can't seem to track down why that is, and it often makes me just go back to digital, hah!

 

[albireo]

Thanks for this. My preferred tool as of late has been Filmomat SmartConvert (w/ vuescan raws) due to being quick for adjustments. I really like how it the UI is based on lab scanners. The lack of historgram is pretty unfortunate. I've mentioned this to the developer. Hopefully they add it in a future release.

You've inspired me to give Nikonscan another go. I have to admit I struggle immensely with consistency no matter what tool I use, be it NLP, Vuescan, Nikonscan, or SmartConvert. Seems like some frames convert really nicely, and some are awful, even when the scene is the same. I can't seem to track down why that is, and it often makes me just go back to digital, hah!

That's interesting about Filmomat. Am I using it wrong? I just don't understand how it can give such (IMO) poor results out of the box. I'd love to find out it's user error at some stage.

Regarding consistency: do you do your own processing? If not, do you trust the lab you're using? I was going absolutely crazy some years ago when my results were all over the place and I couldn't figure out why. I was, at the time, using a small local lab, with very little customer turnover. I switched to another lab, a pro lab with huge customer base, and a clear policy for chemistry replacement. It was night and day better.

If you outsource your development, make sure you rely on someone who always uses fresh chemistry and exact dev times, as these two factors can add a lot of complications when inverting the scans IME.

 

[radialMelt]

That's interesting about Filmomat. Am I using it wrong? I just don't understand how it can give such results out of the box. I'd love to find out it's user error at some stage.

Regarding consistency: do you do your own processing? If not, do you trust the lab you're using? I was going absolutely crazy some years ago when my results were all over the place and I couldn't figure out why. I was, at the time, using a small local lab, with very little customer turnover. I switched to another lab, a pro lab with huge customer base, and a clear policy for chemistry replacement. It was night and day better.

If you outsource your development, make sure you rely on someone who always uses fresh chemistry and exact dev times, as these two factors can add a lot of complications when inverting the scans IME.

Yeah, I process it all myself and this is my first thought as well, sadly.

 

[radialMelt]

That's interesting about Filmomat. Am I using it wrong? I just don't understand how it can give such (IMO) poor results out of the box. I'd love to find out it's user error at some stage.

It just occurred to me... did you explore the 3 different presets available in the Presets menu? Load in your images, and give the "Flat" preset a try (by selecting it from the menu if that wasn't obvious). It will apply the preset to all your currently loaded images. It's a much less contrasty starting point.

And I agree, the app desperately needs a histogram.

 

[albireo]

It just occurred to me... did you explore the 3 different presets available in the Presets menu? Load in your images, and give the "Flat" preset a try (by selecting it from the menu if that wasn't obvious). It will apply the preset to all your currently loaded images. It's a much less contrasty starting point.

Hi Radial, yes I had tried the 'flat' preset, but honestly didn't find it much better. It's probably a better starting point than the normal preset though, if one wants to play with the settings.

Kodak Gold 200 35mm

Filmomat Smartconvert - normal preset

Filmomat Smartconvert - flat preset

Nikonscan

 

[radialMelt]

I mean, I wouldn't hesitate cranking up the density to get the highlights within range. Of course you risk clipping the blacks... I do wish they would add a histogram...

 

[albireo]

I mean, I wouldn't hesitate cranking up the density to get the highlights within range. Of course you risk clipping the blacks... I do wish they would add a histogram...

Colorperfect has it and much more. I'd definitely look into it if you haven't!

 

[radialMelt]

Colorperfect has it and much more. I'd definitely look into it if you haven't!

You've inspired me to give it a try. The built in film profiles are definitely helpful so far at getting me to a good starting point. I don't see a histogram though, and I have to say this software is not very intuitive.

Where is the histogram?

 

[albireo]

You've inspired me to give it a try. The built in film profiles are definitely helpful so far at getting me to a good starting point. I don't see a histogram though, and I have to say this software is not very intuitive.

Where is the histogram?

I was kind of incorrect. The histogram can be manipulated, but you don't see the full distribution: you are given control over where to place the signal at the tails (shadows + highlights) and you can also tweak the 10 zones one by one.

Check out the youtube tutorials made by the guy behind it for an explanation.

But yes I will agree - its interface is very counterintuitive.

 

[Steven Lee]

@albireo I agree with your conclusion: based on the samples you've shared, Filmomat is the worst. Nikonscan has consistency issues, it added heavy color cast to some images. So it is a tie between Vuescan and Colorperfect in my book.

However, my impressions are based on your samples.

When I evaluated Filmomat Smart Convert using my equipment and my films, it blew my mind. I was converting RAW files from Sony A7R IV. Interestingly, when I converted those RAW files to 16-bit linear TIFFs and then fed those TIFFs to SmartConvert, the results were similar to yours. I have no idea why that is. Maybe - and I'm speculating here - it includes an open source RAW library which is bundled with color profiles for popular digital cameras, while vanilla TIFFs don't have that data.

 

[radialMelt]

I have been playing with ColorPerfect for a bit now and it seems to provide pretty good results. I like what I am seeing. Workflow-wise however, I will need to do some research. This doesn't feel like it will be very good for processing a bunch of images in succession (in terms of time investment I mean). Apps like NLP and SmartConvert have a lot of provisions to help save the user time when converting entire rolls.

 

[brbo]

Does ColorPerfect need that if you know how to do PS actions?

 

[albireo]

When I evaluated Filmomat Smart Convert using my equipment and my films, it blew my mind. I was converting RAW files from Sony A7R IV. Interestingly, when I converted those RAW files to 16-bit linear TIFFs and then fed those TIFFs to SmartConvert, the results were similar to yours. I have no idea why that is. Maybe - and I'm speculating here - it includes an open source RAW library which is bundled with color profiles for popular digital cameras, while vanilla TIFFs don't have that data.

That's an interesting possibility. Sounds like it might be a tool optimised to work with digital cameras raw files and not with film scanner raw linear files.

If that's the case, I wish this was indicated on the product page, so that people with film scanners would be able to know results are not going to be satifsfactory.

@albireo I agree with your conclusion: based on the samples you've shared, Filmomat is the worst. Nikonscan has consistency issues, it added heavy color cast to some images. So it is a tie between Vuescan and Colorperfect in my book.

err... that's not how it works. You should have detected and reported the 'heavy color casts' before I revealed the labels. Your judgement is now influenced by any implicit bias linked to you now knowing the labels

However, my impressions are based on your samples.

Exactly. I was in fact hesitant in posting samples. The purpose of my original post was not to find a consensus or try to convince anyone there is an 'absolute' best - it was to suggest Nikonscan gives results I personally really favour, and to find out if other people have found the same when comparing it to different tools, and to ask any experts out there if the signature it imparts to scans (because I find it does impart a consistent signature, based on thousands of scans performed, and in spite of occasional mishaps) is there by design.

In general I would not have expected the fun image comparison to return an absolute 'winner', as people use different monitors, do use/do not use calibration, and have different preferences.

But thanks for chipping in, and I agree with you - this was useful if anything to revisit the humble, built-in color inversion performed by Vuescan, which does a really really good job. Kudos to Mr Hamrick!