I’ve used Richard Photo Lab for a while but decided to give scanning a try myself (I know, the hubris).
I’m using NLP 2.2.0 with Lightroom Classic, scanned with a Prime Film XAs scanner with VueScan 9.7.56. I followed the instructions to scan as raw DNG, applied NLP v2.1 profile per recommendation, and here are the results.
Please see this Dropbox folder, (new users are not allowed to post more than one image or link more than once ), which includes
Original Raw
Vanilla conversion (Noritsu, level 5 pre-sat, auto-neutral. The rest I don’t remember since I no longer can open the free trial…also don’t mind the watermark, I’m still trying out everything.)
My edit
Lab scan for comparison
Notice that in the lab’s scan, the bike and the dark interior on the top left corner are interpreted as red vs. brown, whereas in the NLP scan, both are on the red channel. So when I bump up the bike’s red, the door becomes super red. Same thing with the man’s blue shirt to the right, to correct for the blue cast on the pavement, I had to desaturate the blues, but that man shirt also got crushed in mine, whereas RPL’s maintained that pretty well.
What might’ve caused this? Is this more because of my lack of editing skills, or should I try a different scanner emulation?
Appreciate your help, I’m new to editing in general!
Changing pre-saturation and scanner emulation can provide conversions that differ noticeably. I find that not all scanner emulations give me the results that are close to what I want - also depending on the negative (age, film brand and type, processing etc.) and the settings on the second tab of NLP.
One way to find out what serves you best, is to try different settings. A systematic approach can help to find the way and document it for later reference. Example: Set of “Basic” conversions
rename each copy to reflect the scanner emulation you’re gonna use on it.
create five virtual copies, one for each pre-saturation value
convert each VC with settings according to filename and VC copy number (leave the master alone)
→ you’ll get a grid of images, one row per scanner emulation, one (additional) column per pre-saturation settings. Note that I also color coded the images to add sorting/filtering options.
evaluate results
Once you have chosen the favourite first-tab-settings, you can continue to experiment with second-tab settings. Seems like a lot of work, but it helps to learn what NLP does and how.
Caveat: While many negatives can be converted to your liking with the settings chosen in the trials as outlined above, others might need completely different settings. Most of my negatives were developed where I was at that time which means that there is a lot of variation, even though I mostly used the same film brand and type.
Thank you!
I systematically tried and unfortunately I haven’t found a single combination of settings that eliminated the crushed blue shirt/doorframe problem. I wonder if this is a limitation of NLP vs. Noritsu…
Had a look and go at your file(s) and noticed the following
The scan has a vuescan watermark, which has black and white extremes, which cam throw off NLP.
The scan is very bright. If it were a camera scan, I’d expose it less by 1/3 - 2/3 stops.
Tried a conversion with the following parameters
First Tab: Basic, Presaturation=5
Second Tab: LAB - Soft, WB=Auto-Warm
Painted a few sloppy local adjustments (pullover, bike)
All in all, it looks like a shot that is difficult to convert, due to whatever reason.
I’d try with a tiff scan done with SW that does NOT put a watermark on the image.
NLP has a “Border Buffer”, which tells NLP to ignore the image borders.
If border buffer is set to 0, then the film rebate should be cropped off.
→ we can always un-crop the converted image
If images have not been cropped, border buffer should be set to >0.
If the original negatives have greatly overexposed areas, conversions can be improved if the overexposed areas are cropped off. See this post for more. Beware to not crop in a way that changes the mix of colours too much because it can change results.
Upper row: Artificial image, negative and versions with crop and conversion
Lower row: Uncropped versions of the images above
We can see differences in the results depending on how I cropped the images.
Nevertheless, NLP does a fairly good job under these conditions. Thanks, @nate.
Thanks for trying my files. I didn’t want to pay for VueScan yet until I was sure that I’m going to scan things myself. Just curious, how would the white watermarks throw off the color conversation? (I have another image that also has some crushed blues…it seems to be a pattern)
I use the Pacific Image PrimeFilm XAs btw, if that helps.
How exactly NLP is thrown off, I cannot say for sure, but Nate has mentioned it in the post I linked above. I suppose that NLP looks for brightest and darkest areas which are then translated into black and white. Then, the tone curves are fit between the two extremes. Slight repositioning of a tone curve can change the outcome considerably. You can try it by changing the r, g and b tone curves from / to \ and then drag the upper and lower edges inwards…for a crude reversal.
The tone curves can look like this example (not taken from your shot)
You may not see that the channels are clipping in the current version of Negative Lab Pro, because Lightroom is trying to compress the clipped areas. New version of Negative Lab Pro (v2.3) will override the clipping compression in Lightroom to give more accurate representation in highlights. So here with that clipping compression overridden you can see that you have clipping the red and yellow channels in your original scan. The only fix for this would be to rescan.
Nate, you’re amazing! I used “Lock Exposure” and set the RGB exposure to 1.0 as you suggested, viola, the colors are back! I didn’t understand what “RGB Exposure” setting was supposed to be on VueScan and didn’t dare to change it after clicked on “Lock Exposure” per your site instructions, do you think maybe others might find it helpful too to put this on the site?
I may have to go back and revist the instructions… there are a lot of variables at play… in some cases, like very thin negatives, there is probably still some benefit to locking the exposure… basically you don’t want to be overexposed to the point of clipping, but you also don’t want to have too little information captured.
Thanks, yea I think with 2.3 it’d be easier to tell when it’s clipped, I think maybe a little tip in the guide like “remember to check your RGB exposure” would suffice.
Thanks again!
While I have you…have you noticed any quality difference between SilverFast “real raw” scans vs. VueScan’s?
I’ve seen some really great conversions from both, so I think a lot of it will come down to user preference…
Vuescan has the benefit of including the IR Dust Removal already included in the RAW, but there seems to be a lot of variance in how well Vuescan removes dust, depending on the scanner model. I also personally find Vuescan a bit easier to use than Silverfast, but for people who already have Silverfast and know how to use it, Silverfast is good too.
), which includes
