[Questions] How to choose a DCP profile (if at all)

Hi! (First topic, there.)

For years, I only used a basic compact Panasonic DMC-TZ60 (aka DMC-ZS40), without caring about color profiles and such. As they say, sometimes ignorance is bliss: now I upgraded to a Canon EOS R6 Mark II, and I read a few things about DCP profiles and such, because having an available profile was cited as an important criterion for raw files support in RawTherapee.

:package: I have:

  • A RawTherapee-bundled profile meant for the R6 (Mark I, not II), brutally downloaded from rtdata/dcpprofiles/ in their GitHub repository.
  • Profiles from Adobe DNG Converter:
    • Canon EOS R6 Mark II Adobe Standard
    • Canon EOS R6 Mark II Camera Faithful / Landscape / Monochrome / Neutral / Portrait / Standard

I expected differences between those to be almost negligible, but:

  • They actually yield fairly different results.
  • Most seem to have their own pros and cons.

Here are some examples (yeah, not award-winning pictures, sorry; just me trying to get the hang of the new camera):

Overall, the profiles from Adobe DNG Converter look dull to me. Weirdly enough, sometimes I even prefer the RT-bundled profile meant for the Mark I, even if it looks almost too vivid in some ways.

After a long while spent staring at different RawTherapee snapshots, desperately trying to figure out some kind of rule to decide, I only ended up with more questions. :neutral_face: Here they are:

  1. Should I select a DCP profile early in the “derawtisation” process, or rather after some settings (exposition, details, whatever)? I’d say “early” but I’m not sure of anything anymore.
  2. When applying a DCP profile (and basically just that), should I expect an immediately pleasant-looking image, or is it advisable to aim for a natural-looking picture – even if it looks quite dull at first, like the Adobe examples above – and then count on the rest of the RawTherapee settings (expo, details, colors…) to “fix” the apparent dullness?
  3. Should I ever fiddle with the “Illuminant” setting in the DCP section of RawTherapee, or leave it on “Interpolated”?
  4. Is it blasphemy to choose no DCP profile and leave RawTherapee on the “Camera standard” radio button rather than “Custom”? (NB: At the moment, there is no bundled R6m2 profile, so “Camera standard” probably builds some makeshift profile instead.)
    • Perhaps this could lead to strange surprises if, later on, I process the picture with a new RawTherapee version that does have a bundled R6m2 profile? Like, it would suddenly apply that new profile and that would change the result significantly? This would be annoying, as I sometimes use the RT CLI on old pp3 files.
  5. Should my DCP choice for a specific picture be influenced by which “Picture style” was selected in the camera at the time of shooting? Their names match the Adobe Camera profiles. I think I read that these are non-destructive, like just some kind of metadata, rather than something directly affecting the recorded values per se.
    • I shot the flowers above with the Standard style and the landscape with neutral-but-with-sharpness-strength-set-to-1-instead-of-0 (which I’ve been keeping for all pictures since Day 2, basically – I’ve read somewhere that it allows to get a more realistic view of the available dynamic range or something).
  6. How important is it to be consistent in the DCP choice across pictures? Can I just switch depending on the subject, kinda like I’ve been doing between the “Standard” and “Film-like” tone curve modes? (Wait… could that have been a mistake as well?)
  7. Not 100% related but still color-related and weird: RT 5.10 seems to have a new White Balance checkbox to choose between “Observer 10°” and “Observer 2°”. It has a huge tooltip that says 10° is the default, but every picture I open (for both cameras) yields 2° (i.e., unchecked box) until I manually check that box.
  8. Do you have any tip or heuristic to share on how to pick a profile for any given picture without having to spend hours squinting at 4–6 RT snapshots?
  9. Are there, in my list, some profiles that I should almost never use anyway?

This is bugging me so much that it led me to finally sign up on that website. I feel like a toddler having to re-learn the alphabet all over again. :sweat_smile: (And I was not that experimented to begin with.) I already have a bad habit of wasting too much time on overkill tiny changes in RawTherapee, so I hope the DCP profiles won’t add yet another time-consuming (and not-so-fun) layer to my process. :laughing:

Thanks to anyone who will take the time to answer at least some of those questions. And thanks for existing already, ’cause all those posts were very valuable while desperately trying to pick a new camera a few weeks ago. :bowing_woman:

Welcome to the forum. This may not answer any of your specific questions, but I’ll give my perspective on camera profiles (@Entropy512 is chuckling now…)

Camera profiles in my perspective should be about providing the color information to allow the transform of the camera data to color gamuts for display and print. Cameras have a spectral sensitivity far beyond any current rendition media, so the camera data eventually needs to be “scrunched up” to fit the ability of a rendition medium to display it. Most cameras don’t put this information in the metadata, so camera profiles are enlisted to do that.

Camera profiles, to my taste, shouldn’t themselves specify any other modifications, either to tone or color. Adobe’s DCPs offer variations to do just that, as well as go beyond to modify tone and/or color to make a particular “look”. You can also get third-party DCPs with the same intent. This, I think, is part of your confusion, too many different things to consider in selecting a profile.

Now, DCPs do offer some other things worth considering. For one, if constructed to do such they allow the software to make a color transform tailored to the illuminant at the scene, “dual-illuminant” (they interpolate between the two). Other schemes are “single-illuminant”, usually D65 daylight. I use D65 camera profiles all the time, and I think a good white balance setting drives any distinction into oblivion. But, that’s not a conclusion based on measurement, just my intuition.

Okay, I think I can answer at least a few:

Doesn’t matter when you select it, RT is going to apply it at a certain pre-determined part of the process.

If you go with my predilections, you’ll have a dull image every time. It’s then up to you to choose a tone curve to yield a pleasing result. I think this approach is good in that it lets you first regard what the camera recorded and go from there processing, knowing what you have to work with.

I’ll leave this one to someone more familiar with RT.

Not at all. In my opinion. See above… :laughing:

The camera Picture Control has no effect on the raw data, only the camera-produced JPEGs. That’s the essential beauty of starting with a raw file, you get to determine the processing to make a rendition. That would include choice of “look” DCP, if you choose to ignore my oh-so-sage advice. :crazy_face:

I’d think that one would probably want to use the same “look” DCP for all the pictures of a session, but I can also divine reasons not to. Your choice.

There. I think that’s all I should say for now. I’m not a RT user, by the way, so take the above in that context…

2 Likes

Hi @alicem!

I’m not a DCP expert, but I’ll share what I know for what @ggbutcher hasn’t answered already.

Illuminant: Since Interpolated is the default, that’s usually sufficient. RawPedia says you don’t need to change it under normal circumstances. Now, if you are very concerned about correct colors and the photo illuminant matches one of the available ones in the profile, then you can change it. There shouldn’t be a noticeable change in colors in most cases though.

Camera standard: RawTherapee may recommend settings, but never prescribes them, so do what you want. Camera standard, which applies a simple color matrix, is sufficient in most cases. If a DCP is bundled later on, then the Auto-matched camera profile option becomes available. Because the DCP is only applied when you choose that option, there will be no sudden change to the colors of your existing edits.

Observer: You are right. The default was initially 10 degrees (in the development version), but was later changed to 2 degrees and the tool-tip wasn’t updated. Thanks for pointing this out.

Picking a profile: Personally, I just use the DCP that gives neutral (i.e. accurate) results and rely on the various color tools in RawTherapee to achieve the look I want. For that reason, I have the DCP set in my default profile so I never have to touch the input profile settings.

Profiles to avoid: It’s personal preference for the most part. Do avoid poor quality profiles. The Adobe and RawTherapee ones are good. Also make sure you don’t use DCPs meant for other cameras unless you can verify that the cameras have the exact same color response. I’m guessing that if two cameras have the same color matrix (compare the dcraw_matrix in camconst.json), then they have the same color response. Don’t quote me on that though.

Finally, my advice is to resist the urge to fiddle with every setting. RawTherapee can be very complex. I have seen many instances of people on Pixls (not just for RawTherapee) getting frustrated or spending lots of time trying to understand things that they don’t really need to. Learn things bit by bit as you need them and enjoy your new camera!

2 Likes

Neither am I, but Anders Torger is. If I understand correctly, he wrote a lot, if not all, of the DCP code in RawTherapee, and you can read a LOT about DCPs in his dcamprof documentation. Of particular note is this missive on the workflow behind DCPs:

https://rawtherapee.com/mirror/dcamprof/camera-profiling.html#dng_profiles

The rest of the documentation is a good reference about camera profiles in general.

Based on your snapshots RT “camera standard” looks pretty good. Personally I’d save all the time and efforts trying to understand DCP workflow and just use the standard. I like simplicity that delivers.

There are 3 components to the profile … the tone curve the base and look tables… you should confirm which are active when you use the profile and also if you apply this after and don’t set the automatch tone curve to linear it is adding something on top of what the intended profile might have been designed to generate… so in addition to picking one you have to be careful how you apply it … For example even without the DCP… use the RT defaults and then change the models in the automatch tone curve… I think there are about 6… you will see what a wide range of outcomes you get even from this and so you can imagine if that is somehow piled on to your dcp… I think and others will know better but if you use the dcp you should likely set the auto tone curve to linear and let the adobe curve work with the other two elements of the DCP profile to provide the intended result…

I like to use adobe’s profiles when there are very strong colors in the image because it usually gives a less saturated representation, and you can add saturation to taste.

2 Likes

Thanks a lot everyone! It feels a bit less hazy already.

Doesn’t matter when you select it, RT is going to apply it at a certain pre-determined part of the process.
[…] [F]irst regard what the camera recorded and go from there processing […].
@ggbutcher

Yeah, I kinda knew that RawTherapee applied the tools in a predetermined order; my question was more about what’s convenient for the user (like “Tone mapping” that uses so much CPU that I keep it for the end :laughing: or stuff that may have an impact on other decisions later), so that second sentence answers it: I guess it’s better to choose the DCP early, to avoid a “Gosh I shouldn’t have added so much local contrast, it looks ugly once I use that profile” effect, etc.

I generally fiddle with, in order:

Expand to see boring details
  1. Hot pixel filter.
  2. The non-tone-curve part of the “Exposure” expandable panel of the “Exposure” tab.
  3. White balance.
  4. Shadows/Highlights.
  5. Tone curve (multiple painful iterations, step by step).
  6. Shadows/Highlights again just in case (fine-tuning).
  7. Optional white balance fine-tuning.
  8. Vignette and Graduated filters, if needed.
  9. The other tabs, in order.
  10. L*a*b* Adjustments.
  11. Tone mapping + -3–5 to Lightness and Contrast in L*a*b* Adjustments because for some reason I get weird results otherwise.

This is merely the fruit of some trial-and-error, and may be stupid in some regards. Unless I load stuff from a pp3 from the same series, I need 30–120 minutes per picture. And that was before I discovered (yesterday) that Color Toning could be applied selectively on a mask. :sweat_smile: Things will only get worse from there.

I think I’ll insert DCP profiles at the very beginning of that list.

If a DCP is bundled later on, then the Auto-matched camera profile option becomes available. Because the DCP is only applied when you choose that option, there will be no sudden change to the colors of your existing edits.
@Lawrence37

Oh, I re-read the tooltip. Indeed. There’s the word “embedded” in it, but it’s about stuff from the DNG, so unrelated to profiles in RT. Thanks.

The default was initially 10 degrees (in the development version), but was later changed to 2 degrees […].
@Lawrence37

This explains that. :laughing: But is the “In a majority of cases 10° will be a more relevant choice” part of that same tooltip wrong as well, then? :thinking: Should I keep spamming that checkbox for every picture or leave it be? Well, that generally has close to no effect, but still.

[…] the tone curve the base and look tables […] if you apply this after and don’t set the automatch tone curve to linear it is adding something on top of what the intended profile might have been designed to generate […]
@priort

  • I should indeed have clarified that, for all my examples, the tone curve box was unchecked, and the others checked (default).
  • Oh, thanks to you I finally understand why checking that DCP tone curve box was leading to results with absurdly high contrasts! It was basically combining two redundant curves?!
  • Hum, using the DCP curve + linear RT tone curve seems to give something significantly darker. Not sure It’s a good idea (I guess there’s a reason why the DCP curve is unchecked by default).
    • Oh but if I then hit the “Auto-Matched tone Curve” button after having changed those settings, I can converge towards nearly identical results (between “DCP curve unchecked” and “checked”). That’s interesting (but gives even more possibilities :laughing:).
    • I think I prefer leaving the DCP one off, to get a better visibility of my curve overall, rather than some black-box approach involving a hidden secondary curve.

Thanks again for having read my lengthy post. I may have gone a little bit overboard with that post, but I hope that may help other newbies as well in the future!

And a big thanks to the RT devs. I tried Darktable first and its interface almost made me give up on raws altogether. You even released 5.10 (with more R6m2 support) literally the day before my purchase. :laughing:

3 Likes

According to Jacques @jdc, the tool-tip should read

In a majority of cases Observer 2° (default) will be the best choice.
To avoid a (rare) drift of the colors due to the choice Observer 2° - probably due to the conversion matrix - Observer 10° must be selected.

Source: Update white balance observer tooltip · Issue #6970 · Beep6581/RawTherapee · GitHub

Thanks! So I was doing it backwards. :sweat_smile:
If I may: Perhaps it would be less ambiguous phrased that way (if I got that correctly):

In the rare occurrence where a drift of the colors is noticed with “Observer 2°” – generally due to the conversion matrix –, “Observer 10°” must be selected instead.

But I’m not a native speaker either, so the only thing I can vouch for with any degree of certainty is the replacement of hyphens with en dashes. :rofl:

Note that, at least in theory, you should be able to get neutral results from any DCP profile by disabling the appropriate checkboxes in RT. I’d say exactly which ones, but my local build of RT is broken at the moment (Hangs immediately on startup) and I’m troubleshooting why. I probably need to completely nuke build/ instead of doing “make clean; cmake .; make -j8” in build/

It’s whatever enables the LookTable in the DCP - which I believe shouldn’t be enabled by default - see my comments in DCP profile size inflated with redundant data · Issue #6467 · Beep6581/RawTherapee · GitHub - EDIT: Clean rebuild fixed my RT install, it’s “Look Table” in the options. It’s my opinion that if you have Tone Curve disabled, then Look Table and Baseline Exposure should also be disabled, since these two are tied to the Tone Curve and only match with the default RT “Standard Film Curve” if the user doesn’t touch it.

I say “in theory” because some “look” profiles might put inappropriate things in some of the LUTs that are not intended for “looks”. Bundled Adobe profiles should at least be “correct” in this manner.

Some of the newer DCP profiles in RT completely omit the tone curve and LookTable as they’re not really consistent with RT’s design in terms of implementing the tone curve in a user-defined manner.

Obviously, it’s best that if you don’t know exactly how your profile was generated and trust that process (e.g. I trust any profile generated by the RT team, other than some REALLY outdated ones that ideally might need to be redone), you get a ColorChecker and follow RT’s documentation on generating your own profile with dcamprof. :slight_smile:

This is probably partly because Adobe does similar things to dcamprof with “pre-compression” followed by a 2.5D LUT - see DCamProf

Thank you. This is way above my level, but if I try to extract the main stuff for my use-case…

Since the only profiles I have, right now, for my exact camera model are those from Adobe, and since I’m not fond of using their tone curves (for now), I should uncheck all the options? Well… Perhaps not all: some cannot be unchecked. But uncheck the profile’s curve and the look table?

The result looks a bit surprising to me. It gets nearly identical to what I get with “Input Profile → No profile”, and I get some red-ish color cast. Not sure those are good signs:

I don’t think I would trust a profile that I would have made myself, given my lack of knowledge. :rofl:

It’s not too hard, most challenging part is shooting the target without glare. Just a matter of the right angle…

dcamprof, while a command-line program, does a good job making either ICC or DCP camera profiles. It’ll also report the deltaE (color difference) for each patch, what the profile does vs. the reference color, so you can easily iterate the process until, like SpaceX, it doesn’t explode so much… :crazy_face:

Have a look at this video and any others you find on the adobe profile editor…its quite an easy tool to use and you can use it visually to modify the color and tone elements of your DCP…

Robin is showing it with LR but its the same for RT just create the profile for use… This might work for tweaking your visual style …

I’m using Linux so I’d rather avoid having to deal with Windows-or-Mac-specific Adobe apps. :laughing: I’m used to command-line stuff, so if I really need to build a profile one day dcamprof should do the trick. For now: wouldn’t that be overkill for someone with so little experience? The monitor I use RawTherapee on isn’t even calibrated, for example. :sweat_smile:
I think I’ll first try to get the hang of the pre-built profiles, rather than jumping into yet another rabbit hole so soon. :smile:

The “Look table or not Look table” debate is more or less the last thing preventing me from going wild on my new camera’s raws, at this point.
All the Adobe profiles suddenly make it look like there’s been a drought on my landscape if I uncheck the Look table (and that does seem trivially fixable via white balance alone). I also noticed that the Monochrome profile needs that Look table to “monochromify” the picture. So I’d say I need to keep it enabled, but that seems to contradict the pieces of advice from @Entropy512 (unless I misinterpreted those, which is entirely possible). :thinking:

RawPedia says:

Disabling individual DCP elements are considered advanced user settings, normally you would leave this on.

… but that confuses me a little, as the profile’s tone curve tends to be disabled by default. I suppose leaving that off does not count as an “advanced user” choice?

I ended up using exiftool -b -JpgFromRaw -w _thumb.jpg ./*.CR3 just to double-check that that landscape was indeed green-ish. :rofl: This is driving me insane.

I suspect this is likely because I think RT starts with the low iso auto tone curve as the default so if you also enabled the Adobe one you would be doubling up…

1 Like

You can over think it and bat it around all over the place. If you don’t calibrate your camera and monitor then in a way you are already working in the abstract so worrying about the look table etc are kind of just a hit and miss enterprise.

I think the advice to uncheck everything was to give you a neutral start and not advice on how to use the profile…

RT has quite a few options. Some people take comfort in the Adobe look and embrace that as what the image should look like and others do not.

I would expect the profile is designed to work with the adobe tone curve and manage any hue shifts introduced by that tone mapping curve. Sometimes people are looking to start out with the look of the jpg and others want to be free of that so I guess you could decide which of those reflects your preference and then try all the options Adobe and RT provide for color and see what gives you what you are looking for….

Oh, indeed, and if they automatically “linearized” the RT tone curve to fix that, people would be kinda surprised I guess.

I guess so. I was trying to find some middle ground, but there’s something weird going on with my white balance – it might be why I’m struggling so much over nothing. I was actually writing about that when you posted:

Here are my observations (taking the exiftool-extracted JPGs as references):
Regardless of the picture I load, the white balance in RT is initially set to 5257 (with a Tint at 1.043), which is often too high, especially for indoor pictures. That makes DCP selection hard, as the output then doesn’t make much sense regardless of my DCP choice. I initially suspected something like that old bug, but then noticed that the metadata was lame to begin with: All the raws of my new Canon have:

White Balance       : Auto (ambience priority)
White Balance Red   : 0
White Balance Blue  : 0
Color Temperature   : 5200

… while the old Panasonic raws have ever-changing non-zero Blue Balance and Red Balance values (without the word “White” but I guess that’s the equivalent for that brand), as well as a Color Temp Kelvin that also greatly varies. :thinking:

I thought that it was caused by the fact that I blindly followed an online advice saying:

Shooting Menu, Tab 4: White balance: AWB W (White) (I seldom use another white balance setting while shooting, though I often adjust white balance modestly during post-processing)

… so I tried reverting this to AWB-without-the-last-W (the default). No luck: same result, same constant values, both in the metadata and in RT:

bear

Also note that, strangely enough, the metadata says “Auto (ambience priority)” when using what the camera calls “Auto (white priority)”, and just “Auto” when using what the camera calls “Auto (ambience priority)”. What the hell.

Even the info displayed by the camera itself when reviewing the pictures show a white balance diagram in which the dot is desperately stuck exactly at the center of two color axes, so it’s not exiftool’s fault either. And yet, the pictures’ previews and embedded JPGs look fine.

Obviously, there are things that I don’t understand regarding white balance.

  • Why would my cheap old compact camera’s “Auto” white balance mode really adapt to the shooting conditions and produce ever-changing and very helpful values while the R6m2 would not?
  • Did I mess up with some settings? I couldn’t find anything suspicious.
  • Am I supposed to use some sort of RT trickery to deduce the WB from the embedded thumbnail or something?
  • Why can’t I find anything even remotely relevant by googling stuff like canon why all pictures same temperature even in auto?

I doubt it’s related, but FYI I’ve been using the “Flexible Value” (“Fv”) camera mode. Edit: Checked in a later post. Switching modes does not help.

I feel even sillier than at the beginning, ’cause the culprit never was DCP profiles, it seems. And I should be asleep right now instead of banging my head on this. :laughing:

Been there, done that, multiple times…

Here’s a perspective to consider; it’s one where I finally came to terms with all you’re wrestling with:

I think you have the right idea, get a true ‘neutral’ rendition. To do that you want a minimum of processing, and that processing to be well-understood. In RawTherapee, I think that’s the Neutral processing profile, with the camconst.json camera primaries. For white balance, you want to use the Pick tool on a white or neutral patch of the image. Dump all that DCP/temp-tint stuff. That way, you get to consider the image in as close to what the camera was calibrated to capture regarding colors. Then, you can mess with the other things and consider what you see in terms of difference to the original linear/min-processed image.

Determining what’s going on in an image really can only be done if you know the starting point. This perspective has kept me out of the bars…

1 Like