Reddish Results in DC Mode

[KERICH]

Hi everyone,

I’ve been experimenting with DC mode in ColorPerfect 3, but I’m not getting the results I expected — all the processed images come out with a strong red cast. I’d really appreciate any tips or feedback from the community about what might be going wrong.

Here’s exactly what I’ve been doing:
1. Export a linear TIFF via MakeTiff (with the working color space set to ProPhoto RGB).
2. Open the TIFF in Photoshop and crop it so only the film negative remains.
3. Launch ColorPerfect, choose ColorNeg, and switch to DC mode.

Even after carefully following these steps, the output still looks too red.
Has anyone else run into this issue? Am I missing an important step or setting?

For reference: when digitizing negatives, I always expose about +2 EV to the right to ensure the best possible signal-to-noise ratio.
This particular negative is Kodak Portra 160.
I’ve uploaded the RAW file so everyone can download it and give it a try — I’d love to see how you process it and whether your results look different.

Any advice, examples, or experience with DC mode would be greatly appreciated!

Thanks in advance

[Full image link - opens in new tab]

[robyferrero]

In the meantime, I can tell you that the film chosen for characterizing the negative
is China Lucky - Luckycolor superGBR400.

Then go through FilmType - Subtype - FilmGamma, even repeatedly, first one, then the other, and then the other, and if desired, repeat the steps if it's still not convincing.
With this process, you'll find your preferred characterization.

From there, proceed with the other tools, preferably starting with FilmicRelight.
It's possible, and likely, that you can also characterize it with other film characterizations. You can also find the characterization just with FilmType - SubType - FilmGamma, without going through the characterization of a single film.

Film characterization is primarily designed to characterize an entire reel, not a single frame, but nothing's stopping you from starting there. Also, because you're characterizing a frame for the first time, which could mean you're also characterizing all subsequent frames of that film.

[robyferrero]

The FilmType - SubType - FilmGamma adjustments can be made with Maximum Saturation, to simplify the process. Once this is done, return to Nominal Saturation.
Here's a detailed description on the topic.

https://www.colorperfect.com/filmtype_s ... ml?lang=en

[C.Oldendorf]

I was tempted to go at it with FilmType / SubType / FilmGamma but stayed with Portra 160NC as shown.

[Full image link - opens in new tab]

This is just a crop from the above - to emphasize areas that work well.

[Full image link - opens in new tab]

First and foremost, what we see is a color balance issue.
ColorPerfect takes the brightest highlight it can find initially and assumes that to be in reference to the light source.
If the brightest object is blue sky, that can fail; if it is a yellow flower, it will fail.
The challenge comes from the densest region of the film. It is possible that the shoulder region effects I discussed with Alexis yesterday play into it.

To color balance find something that should be close to neutral like one of the rocks in the stream, not all of them but that was what I used.

What I did can be loaded as a *.CPMetaN file into CP3.

In the above, I let the sky blow out. We could have compressed it back in, too.

[Full image link - opens in new tab]

Whether we like its color or whether that needs to be tweaked in creative editing is a decision for later. (I would in PS I guess, but I think I like the blown version better)

[Full image link - opens in new tab]

[robyferrero]

From the tests I'm doing, it seems the best result can be achieved without film characterization. So leave it on user, and adjust everything from FilmType - SubType - FilmGamma.
Keep in mind that I'm a user like you, not a technician. We're here to try, experiment, learn, and help each other out.

Then, fortunately, Christoph comes along with the best example.

[C.Oldendorf]

Then, fortunately, Christoph comes along with the best example.

Only because I helped build the system and then had to take over carrying it forward does not mean that individual images are any less of an adventure to me than they are to others. I do not think there is such a thing as “the best.” What there is, is inspiration on what to try. I do not intend to establish any dogma about ColorPerfect’s use or imaging in general. My aim is to offer control and opportunity, from which hundreds of edits become possible. Additionally, with DC the required transforms are of an extreme nature. It is a small miracle that it works as well as it does. It will never be a scanner, but it can come close now, and that is huge.

[robyferrero]

Of course, there's no such thing as "best," but there is "works."
And so far, when it comes to putting in a little more effort, your images work.
Personally, even if I add my own interpretation, it's you who inspires me, because I see that you know how to extract everything from the file.

[KERICH]

Thanks a lot, Christoph!
I had never paid attention to the Film Gamma feature before — using it to correct the color cast is really powerful. Your explanation and the sample you shared helped me understand how to balance the image so much better.

[KERICH]

Based on what you added on Xiaohongshu I’d like to ask why does ColorPerfect use the brightest area as the reference for the light source
The brightest part of an image is often not neutral gray
Does that mean I should place a gray card in the frame when photographing negatives to get the most accurate white balance
And if every time I have to find an area that’s close to neutral in the image, isn’t that kind of a matter of luck? What if there’s no neutral gray area at all — wouldn’t that be a real problem?

[C.Oldendorf]

Based on what you added on Xiaohongshu I’d like to ask why does ColorPerfect use the brightest area as the reference for the light source? The brightest part of an image is often not neutral gray

Maybe the wording was unfortunate. "Use" suggests it’s simply saying “that must be the light source,” but the reality is more nuanced. Our research showed that in the vast majority of cases this provides the best initial estimate possible. It’s really a matter of weighting, and most of the time it works well (except in cases like a macro shot of a yellow flower, which can initially give a blue image).

It’s also user-controllable. The AutoColor feature adjusts those weights, and if you take your Sakura image, click into AutoColor and increase it, you’ll get a nearly correct color balance around the 850 range of the function.
That is essentially the same technology, only with different weightings, and it’s quite close to what some digital camera systems use in their automatic white balance modes.

Does that mean I should place a gray card in the frame when photographing negatives to get the most accurate white balance?
And if every time I have to find an area that’s close to neutral in the image, isn’t that kind of a matter of luck? What if there’s no neutral gray area at all — wouldn’t that be a real problem?

It’s the same problem photographers have always faced in the analog darkroom. What really matters is consistency in digitizing: using the same light source at the same intensity and distance, which I assume is the case with your university’s Phase One IQ4 150MP setup. If you fix the back to manual exposure and keep the aperture constant after establishing how bright film base should be (since nothing in a negative can be brighter than the film base), then all frames from a given roll of 35mm or 120 film will align consistently. That level of rigor is sufficient that you can even stitch multiple close-up captures of a single 4×5 negative and expect them to match seamlessly — with a system like the IQ4 you could probably reach the 400-megapixel range from four stitched frames, which is something worth trying in the future.

For color balancing, however, you don’t actually need that degree of rigor in exposure, because exposure changes only affect relative brightness. A neutral patch at any brightness will still come out neutral when you click on it, so gray balancing remains valid even with exposure variation.
This principle extends even to large-format negatives like 4×5 and 5×7 from the same batch developed together, though with sheet film I often find development variations to be a bigger factor than color balancing itself.
What really matters is using the same light source. On top of that, you also have tools for visual refinement, such as Ring CC, which lets you arrive at the balance you want by eye.

[C.Oldendorf]

KERICH, LEON,

Thank you so much for adding that beautiful second image. That’s a place I’d love to visit in autumn, and it ranks high on my list.

When I opened it in CP3, my first thought was, “this isn’t half bad.”
But then I realized exactly what you’re running into.
The key takeaway turns out to be simple yet important.

This is what I saw when I opened CP on your photo:

[Full image link - opens in new tab]

To make a minimal example, I raised AutoColor a bit higher to 550 as a starting point:

[Full image link - opens in new tab]

However, I thought, why would they see something so different? And then it hit me.

The initial guess at color balance is tied to the initial brightness estimate. On the Options panel, the Lightness threshold defaults to 0.002500, while I had mine set to 0.005.
If I set my threshold to 0.00005 instead, guess what I get:

[Full image link - opens in new tab]

So for now, with an extremely low starting point on brightness, use AutoColor. Or set a higher starting point and use Black to make the image as bright as you’d like.

Actually, I might be inclined to decouple the two, so there’s always a minimum threshold for the initial color balance guess.
As we’ve just seen, things can otherwise go badly. Then again, out of the box with default values everything works fine — it’s up to the user to break things.

[C.Oldendorf]

So that everyone can get an idea of what we are doing here, let's see a display corrected rendition of the negative, our input data

[Full image link - opens in new tab]

Since we have this beautiful thing already, why not give it an actual edit like we mean it.

[Full image link - opens in new tab]

I posted it at full resolution. I did some despeckling, but didn’t get all of it.

There’s something you might want to look at: Newton Rings, lower right corner at full resolution.
They’re produced by a very thin air gap between two transmissive surfaces, and are therefore intrinsic to your repro setup.

I essentially made two versions and blended them. I can go into the “how” — hopefully I’ll still remember when the time comes — but since I saved two *.CPMetaN files to preserve the edits, I’m attaching them as well.

PS: This is Kawaguchiko, right? I've been there in January 2011, as you know...

[robyferrero]

Very nice!
Beautiful image, beautiful color, and beautiful editing.

[KERICH]

Actually, I haven’t changed the Lightness threshold — it’s still at the default value, 0.005, the same as yours.
Although the initial image we see is the same as what you showed — you first thought we were seeing something different, but in fact it’s identical — you consider that acceptable since only a slight adjustment is needed to improve it.
However, this is not the same as what I was used to with the workflow from ColorPerfect 2.0, where the output could be very neutral straight away.
Maybe I’m just being too strict about how neutral the out-of-the-box result should be, but it still feels different from what I was used to.
So the behavior we noticed wasn’t because I set it too low — maybe something else is affecting the result.

Also, I’d like to show you the workflow I used back in the ColorPerfect 2.0 days for handling DSLR copy shots of negatives.
It was the process you once shared with me, and it could produce a very neutral result without any further adjustments.
The only drawback was that sometimes blue skies would shift slightly toward cyan.
I’m planning to make a short video demonstration, showing this workflow in ColorPerfect 2.0, and point out the differences in the results.
I’ll also include several side-by-side comparisons of straight-out-of-CP images from version 2.0 and version 3.0, so you can clearly see how their outputs differ.

[Full image link - opens in new tab]

At that time you suggested performing a film-base white balance step in PerfectRAW.
I find this step very important, because it reduces the influence of the light source and other copy conditions, and helps to standardize the workflow: it always treats the film base as neutral gray.
The base area and the unexposed part of the image are the same — after inversion they become the darkest part of the image — and I believe the white balance there must be neutral.
That’s why I don’t quite understand why ColorPerfect 3.0 doesn’t seem to offer any comparable film-base white balance step, especially since the current white balance approach doesn’t seem to be as accurate.

By the way, the photo was taken at Lake Kawaguchi, specifically at Ubuyagasaki, a spot famous for its beautiful red maple leaves in autumn.

[C.Oldendorf]

KERICH, LEON,

thank you very much for uploading the video — that makes explaining the situation much easier.
I used tiny pink circles in the images to highlight key points.

You are using PerfectRAW in a way we will examine further below. PerfectRAW outputs gamma- or tone-reproduction-curve–encoded data.
Since you configured ProPhoto RGB, PerfectRAW outputs Gamma 1.8 encoding to match that, not linear.

Yet in ColorNeg SC (“Scanner” or “Standard Capture”), you left the G/L button at its default L.
That setting defines the input data as linear, which it is not.

[Full image link - opens in new tab]

OK. In PerfectRAW you do this:

[Full image link - opens in new tab]

As we discussed, that produces Gamma 1.8–encoded image data.

In ColorNeg SC you then treat it as linear input:

[Full image link - opens in new tab]

The correct setting, regardless of how it initially looks, is Gamma 1.8.
This puts us exactly where we would be with ColorNeg DC, without the intermediate PerfectRAW step (which only worked on some cameras).

[Full image link - opens in new tab]

To illustrate what the wrong input gamma does, let me show you. It effectively implies a radical shift in FilmGamma.
We click on FilmType to get to SubType and, with it, FilmGamma:

[Full image link - opens in new tab]

For Portra 160NC, FilmGamma starts at 1.661 per the built-in characterization. The gamma mismatch here is the inverse of 1.8, i.e. 1/1.8 = 0.555.
Multiplying 1.661 × 0.555 = 0.9227 produces the same visual effect as using the wrong input encoding. This is a totally arbitrary shift in Film Gamma.

[Full image link - opens in new tab]

If we don’t apply that arbitrary shift in Film Gamma but instead use AutoColor, we come full circle and end up right where we were in our discussion above but starting from DC mode and without the extra step which has shown it can be a source of misunderstanding.

[Full image link - opens in new tab]

it is totally possible that we need to lower FilmGamma to match development conditions,
but not this much and it would be good to know that we do rather than doing so by accident and in a hidden way.

[robyferrero]

So for digital camera reproductions it would be more suitable to work with Adobe 1998?

In Kerich's video, I notice that he makes some changes, such as excluding BPTails, exposure and white balance, still in PerfectRaw, only after that he switches to ColorNeg.

[AlexisMagni]

Following this thread from afar, but with great interest.

I really liked the solution of being able to neutralize the orange base manually before continue with the conversion, setting this way a solid reference point to move forward.

Would we be able to do the same working with Linear Files? Or does it require using PerfectRAW, and then straight to Color Negative Conversion?

[C.Oldendorf]

So for digital camera reproductions it would be more suitable to work with Adobe 1998?

Personally, I see no drawback in using ProPhoto RGB with ColorPerfect 3.x — other than the fact that it can contain colors your screen won’t display. An Adobe 1998 “superiority” cannot be concluded from the above. That discussion was about a G/L mismatch: gamma-encoded input data versus linear input. If your wide-gamut display is closer to Adobe RGB 1998, then using that would make sense. On 8-bit per channel images Gamma 1.8 is poor and ProPhoto RGB catastrophic, but as long as we have 16-bit data we are fine either way.

In Kerich's video, I notice that he makes some changes, such as excluding BPTails, exposure and white balance, still in PerfectRaw, only after that he switches to ColorNeg.

All of that is outdated; it predates ColorNeg DC. That is why we compare the ColorPerfect 2.x and 3.x workflows. ColorNeg DC now handles all of those steps — and more — automatically.

I really liked the solution of being able to neutralize the orange base manually before continuing with the conversion, setting this way a solid reference point to move forward.

I understand the psychology of that. The infamous orange mask is what countless sources blame for the challenges and failures of color negative conversion — and here you can make it disappear with a click. Only, doing that is not actually needed. In the PerfectRAW intermediate method discussed above, the only reason to do it is to avoid fully saturated pixels and loss of detail in the 15-bit per channel intermediate (in Photoshop what they call 16-bit is in fact 2^15+1 values, for historical reasons). The same automatism that removes the orange mask in ColorNeg DC processes your input in ColorNeg SC or 2.x and sets its own base, independent of your choice. Setting two completely different color balances in PerfectRAW with the above method will still lead to identical starting points.

Would we be able to do the same working with Linear Files? Or does it require using PerfectRAW, and then straight to Color Negative Conversion?

We cannot. Well, in SC mode you can mangle the channels any way you want — if you don’t blow anything, it had no effect before and will have none now. In DC mode, the digital camera input data may only be scaled in brightness as a whole, never in individual channels, or the camera calibrations will fail — more often miserably than subtly. When I put together a higher dynamic range color chart from your shots, I scaled them all to match in brightness first, so that the color channels would remain in the same relative balance.

[AlexisMagni]

Understood and nothing further to add!!

[KERICH]

I understand now — it was entirely the result of an incorrect workflow (even though, by coincidence, the outcome from that mistake looked quite pleasing to me, despite not being technically correct).
That said, I think there’s still some room for ColorPerfect 3 to further refine how it handles color balance, which could make the process even more robust in the future.

[C.Oldendorf]

KERICH, LEON,

First of all, I would like to thank you both for carefully studying what I posted and for engaging so constructively. I am very relieved that I was able to convey what was going on. My core intention was never to prove anyone wrong or to lecture on how things “should” be done, but rather to foster understanding — and to do that, the explanations sometimes must get intricate.

As for the color balance you observed, I believe exposure plays a key role. In the Kawaguchiko image at Mount Fuji, the specular highlights on the lake — essentially a thousand little suns — drove the film to its maximum densities. With the Sakura image, shoulder involvement is a possible explanation, though not yet certain. Since you mention that the same situation occurs with other images as well, it would be helpful to know whether there are cases where it does not occur. I ask because other users of the Phase One IQ4-150 and IQ3-100 Trichromatic backs in repro work do not report this behavior. And just yesterday a fellow user from China shared an overexposed frame, retro-photographed on a Panasonic S5, where the sky showed a color discrepancy that could only be corrected in post-production after ColorPerfect — the same direction as your Sakura example, and meant to illustrate that you are not alone in observing this.

That said, I do think there is some room for ColorPerfect 3 to further refine its initial guess at color balance. As I mentioned, if the threshold gets too low, we can run into trouble. I will look at that, although it appears to be a larger undertaking, and my original plan was to focus on bug-fixing rather than new features. Still, this lies somewhere in between the two.

It is important to emphasize that the initial automatisms are merely guesses — a starting point for what might look pleasing enough to begin with. From there, the user must make their own aesthetic decisions, just as has always been the case in the analog darkroom with color balance filtration and related choices.