Very inconsistent results in the same roll

I find I’m getting wildly fluctuating and inconsistent results (especially color temp) from batch converting the photos from the same roll and the same scan in NLP.

My process is, I import the raw files to LR, then choose color temp based on a blank edge area of the negative, crop, and run through NLP with the default settings.

Some rolls of film come out looking great. And then sometimes, I get this. The first two images come out looking incredibly blue. And then the rest of the images are normal-ish looking:

The blue image settings:

I then tried converting the blue image individually in NLP. It actually works out pretty well:

The same image run through FilmLab and SmartConvert:

I tried throwing the ghastly blue image into SmartConvert and FilmLab, and they both produced relatively reliable results. So am I doing something wrong in NLP?

Here’s the raw file in case anyone wants to have a try: https://1drv.ms/i/c/499d6900d1cab18c/EbRaKUUkM3xChbamyKhKf1wBlXU8JjNto6aUJ6QWPWVMVw?e=4fVWUb

Hi there, I had the same problem when I had ‘roll analysis’ box ticked. Untick this, this try doing all of them again. Worked for me.

Hi,

Yes, something has gone wrong here… it should not make this very blue image.

Let’s figure out what has happened…

  1. Go to the “roll” tab and select “individual analysis.” If this fixes the blue hue, that would probably mean that this image was not captured at the same settings as the other images in the analysis, and thus, the roll analysis will be thrown off.
  2. If that doesn’t fix it, select just this image, then go “convert” and unconvert the image. Select “preview” and adjust the border buffer until ONLY the film is showing (and not any of the surrounding border). Then, click “Convert Negative” and see how the result is.
  3. Then, try changing the WB to either Auto-Warm or Auto-Mix.

BTW, I tried the sample image you provided and didn’t have any issues, but it is a different image than the example you gave.

I do see that you have a lot of direct, unmasked light also hitting the frame in the un-cropped version, which will definitely hurt the quality, and you can see it in the converted image… you will get much better results if you mask out all that direct light from your light pad.

Sorry, I must’ve linked to the wrong raw file.

  1. Go to the “roll” tab and select “individual analysis.” If this fixes the blue hue, that would probably mean that this image was not captured at the same settings as the other images in the analysis, and thus, the roll analysis will be thrown off.

Do you mean set “Analysis” to “This Image Only”? If so, I’ve already done that. I’ve also tried turning on and OFF roll analysis. It doesn’t make too much of a difference.

The only thing which makes a lot of difference is if I convert them individual vs selecting multiple photos and batch converting them. If I batch convert, even with roll analysis off, I get very blue results.

But anyways, it seems to be pretty random. I can’t

Another problem I’m encountering is that I’m getting a very ugly blue vignette in my photos. I had worried this is due to some error in my shooting (I’m using a Canon 100 Macro + Sony A7RIII to digitize). But when I tried to scan with my Epson V700 scanner, it gave more or less the same results.

If I run the same images through the other programs (FlimLab & SmartConvert) the blue vignette is much less pronounced. It’s still there, but tolerable.



@Nate, I didn’t want to post all the raw photos publicly, but I have emailed you a OneDrive link to all of them

I have the impression that the vignette is due to uneven lighting of the original takes. Looks like they were taken with a flash and a wide angle lens.

Most probably it’s in the original negative. Why can’t it be a normal vignette where only the exposure falls off towards the edges, and not have the color cast?

I have some prints which were made at the time (40+ years ago). Although they’re faded and yellowed, they only show a light falloff and no color changes as you move from the center to the edges

If we consider the captures to be taken with mixed light and NLP’s way to balance conversions, some colour cast can be expected. Local adjustments done before converting might help to compensate the cast, but it’s difficult to say how well it will work. Check out the following post on how to try/do it:

I am shooting the Canon EF 100 Macro lens at f/11, adapted with a Metabones IV to my Sony A7R III.

I tried shooting a blank image of the light source. It seems fairly uniform to me. Not much light falloff. I tried your suggestion of using various size and shaped radial filters to make the histogram peak more tightly, but can’t. Any radial adjustments I make only seems to make the histogram wider.

The only way I can seem to make it tighter is by enabling profile correction. Even then, as you can see, the difference is not too large

I tried enabling / disabling profile corrections on the actual photos to little effect

I found some of the prints made from these negatives way back in 1986. I scanned the prints with an Epson V700. Although the prints faded a lot, and there’s a clear red cast over them (easy to correct in LR), the color is uniform. Unlike in the scanned negatives where the colors vary so much.

Left side is from the prints, right side is from negative.

Okay, your backlight seems to be even. What about the negative? Does it show any vignetting or falloff when you look at it with your eyes?

The hardcopy scans show some falloff. So, you’ll have to correct that before you convert. Also, the reflection that shows in NLP output seems originate in your camera scanning setup. Looks like any combination of stray light and reflections.

Do you camera scan in total darkness? Is there any possibility for light to get in between the negative and your lens? Can you add a photo of your scanning setup?

Okay, your backlight seems to be even. What about the negative? Does it show any vignetting or falloff when you look at it with your eyes?

I’m not sure… The negatives are quite dark… I don’t really have much experience with negatives. Never shot them before, and I’m scanning them only only because I have so many old family photos. Anyways, I don’t see any obvious vignetting

The hardcopy scans show some falloff. So, you’ll have to correct that before you convert. So, you’ll have to correct that before you convert.

By “correct that”, do you mean the falloff in the scanned prints? Or in the negatives? Should I be correct any falloff in while still in the color negative stage? If so, it’s kind of hard to gauge the falloff in that mode.

Also, the reflection that shows in NLP output seems originate in your camera scanning setup. Looks like any combination of stray light and reflections. Do you camera scan in total darkness? Is there any possibility for light to get in between the negative and your lens? Can you add a photo of your scanning setup?

Yes I do scan is basically total darkness (lights all turned off, curtains drawn).

I worried that my light wasn’t up to snuff. So I bought two more CRI 98+ light sources today. Unfortunately, they produced more or less the same results, so I guess the light source isn’t an issue.

I’m using a setup from a new company called Re:Film. Their stuff is supposedly based on Negative Supply’s (i.e. ripped off), but is much cheaper at ~$200 USD

Re:Film setup


Trying new light from Godox, rated at CRI 98+…


Left side is using the Re:Film light (before), right side is the Godox light I got today(after)

Please try the following:

  • put your negative on the backlight without any other holders, funnel etc. and
  • cover the light that is not going through the film
  • capture the negative, so that it does NOT fill your sensor (set to e.g. 0.5 or 0.33)
  • post the uncropped negative


Here are the raw files: Microsoft OneDrive

Thanks for the screens. I used one of them to check your situation as seen with DxO PhotoLab (or any other tool and manual settings)

Backlight and exposure:


Lighting seems to be fairly good with a slight hotspot in the area marked with black pixels indicating burnt highlights. Exposure might be a little bit too high, nevertheless, image content was not harmed. Still, I’d probably reduce exposure by maybe 1/3 or 1/2.
→ Bracketing can help to find the optimal exposure, specially with difficult negatives.

The negative(s), manually converted in B&W:


I set development in order to show the light falloff in the original negatives. The falloff is caused by the specific lighting. The light source reflects off the wall in the upper two photo and is hidden in the third image. The falloff can be seen on the wall, easily made visible by stretching micro-contrast and exposure, the banding is due to the 8-bit jpeg format.

Colour vs. B&W
There is not much variety of colours in the photos. No matter if I used NLP (which is happier with a wider variety of colours) or manual settings, output colours were always subdued and the yellow-blue falloff remains visible. In order to make the colours resemble the ones on the hardcopies, I’d probably work on positive copies (using 16 bit TIFF) and tweak colours in Lightroom or any other app. At times - and when memory is more important than accuracy - I convert difficult images in monochrome, e.g. B&W with or without effects like sepia tinting, duotone etc.

Have you tried other programs to invert the negatives scanned by the camera?

I tried FilmLab and it seems to show much less color vignetting than NLP. Of course, the light falloff is still there, as expected. However, my strong preference is to remain in NLP as all my workflow is in LR.

What do you think may be causing NLP to make the color vignette so much more saturated and visible? And is there any (combination of) settings which can counteract this?


Many converters exist out there and some might be better suited for this kind of difficult negative. Whether one of these apps is a cure-all, I can’t say. The question remains whether to stick to NLP, use something else or use apps that best fit the needs.

Using several pps/ways is just a proliferation of effort and so far, I’ve not been ready to go down that alley. I’ll stick to B&W if I can’t get the colours I want…or have a lab make corrected prints. Not really cheap, true, but a few images might be worth it…even if the result comes with moodily subdued colours. We might accept them as “character” or reject them. Our own choice, really.

I have photos on 40+ years old film, some of them convert easily, others are just plain obnoxious. Taken in low, mixed, uneven light and underexposed is the biggest source of headaches as far as I see. Also, some of those negatives were almost impossible with NLP version 2, version 3.0.2 has changed and also improved the output and version 3.1 might be coming soon und deliver the ultimate WOW! If not, I can always resort to B&W or a “vintage look”.