What is required to get a good starting point? (i.e. something close to the JPG from camera)

In the past I’ve asked for feedback regarding the steps to process pictures because by default the RAW image looks very different from the JPG that comes out of the camera.

While reading Darktable 2.4.0 release notes, I noticed the darktable-chart program, read a bit more about it and found @harry_durgin’s video on how to use it to apply a base curve that matches the colors from the camera.

Now, the first step to that process requires a pricey color checker (by the way, any recommendation on what color checker to get is more than welcome!), and since I’m a beginner I would like some input before digging in that direction.

So, if I want an out-of-the-box RAW file that looks close enough to the JPG from the camera:

  • is that color checker + darktable-chart enough? Is there anything else that need to be done to have a good default starting point when editing RAW?
  • do I need to create a preset only once?
  • will shooting in very different conditions render this step useless? (for instance if I shot the color checker on a sunny day at noon as recommended, but then the photo I’m working on is shot at night in a candle-lit room)

I know the JPG from the camera is not necesarily the best output, but I’ve been using the same camera for five years and I think most of the time the JPG output is very good. It’s just that I would like to start from that point and make some adjustments and this is better achieved when using RAW…

Anyway, thanks for your feedback and happy holidays!

Can’t answer the questions since I don’t really go for the OOC jpeg look for starting point (prefer a more neutral starting point but YMMV) but for targets have a looks at http://www.targets.coloraid.de/ have ordered from there before and they seem to be ok.

The in-camera JPEG is one possible result from processing the raw. My camera makes a fairly “flat” JPEG. I can usually find a result that I prefer, so I don’t take the in-camera JPEG as a starting point. My automated process makes a straight sRGB image and a few variations, and I start from these.

However, the JPEG is often interesting. Most of the difference from a straight sRGB result can be made by applying a 1D clut (ie, independent cluts to the three RGB channels). The clut to do this can be calculated from the histograms of the JPEG and sRGB, and is called “histogram matching”. After applying the clut to the sRGB, its histogram will roughly match the JPEG’s histogram, so the images will look very similar. This doesn’t need colour charts or calibration.

I explain the process in Process modules: matching histograms

Using darktable-chart and a color checker will give you a more accurate starting point, colorwise. I don’t think it will put you and closer nor further away from getting a SOOC jpeg look. Note that more accurate does not mean more pleasing nor better looking.

I’d agree with @snibgo, you should be able to meet or surpass the SOOC jpeg pretty easily, by finding a few modules that give a pleasing result.

dt-curve-tool will give you a decent approximation of the sooc jpeg. It won’t give you the same sharpness but colours can get close. It’s good to know your way around a computer before using it though.

Personally, I would edit the Raw in isolation and then compare it with the JPG. If there is anything I like about the JPG that my edit doesn’t have, I would try to incorporate it at that stage. In the end, it is a matter of taste, unless you are using your images for a technical reason.

1 Like

Thanks, it looks like a very interesting tool, I’ll give it a try. I searched a bit and ended on this article from 2013 which mentions that we should use blurry, almost black and white pictures… but if the purpose of this tool is to get accurate color curves, what’s the point of using almost black and white pictures?

Also, the helper script seems to use a list of RAW files (not just one). Does more RAW files mean better accuracy? Do they have to be very different?

If you don’t have a ColorChecker to shoot, you could use the relevant files for your camera from the Imaging-Resource website. It’s not ideal, but should be close enough. I use DCamProf to generate ICC profiles for use in darktable. There are also other options (I wrote about it in this post).

Matching an in-camera jpeg is not a good idea if you’re an ETTR practictioner when shooting because that jpeg will be flat, slightly overexposed and might even have a UniWB applied. Design your own camera profile taking into account both tonality and colours that satisfy your aesthetics.

I stopped altogether doing OOC JPEGs for a different reason: size. I’d already come to the conclusion others describe above about not needing to duplicate the camera JPEG with curves and profiles and such, but I continued to shoot to JPEG for snapshot things like family events.

It was on our recent vacation that I came to a workflow that started with the raw files and batch-processed them to JPEGs of the 800x600 size for posting. These JPEGs are a lot easier to post from dodgy hotel wifi. That also gave me a good “contact sheet” for review, and selection of images for more attention in PP.

I keep our old D50 set to JPEG-small for my family to grab if they need, but if I shoot anything, it’s with the D7000 set to RAW and I batch-process as described above. For most family uses, the small image works just fine.

Oh, my batch processing “secret sauce” is a simple contrast stretch, a “linear” curve with the black and white points moved to the edges of the histogram data. I find that operation deals well with varying approaches to exposure; sometimes I’ll put in a stop or two of over-expose as a goofball attempt at ETTR, and the batch contrast-stretch produces consistent JPEGS for them as well as the so-called “normally exposed” shots.

If you’re shooting Nikon you can get the jpg “profiles” from Nikon software. They match: http://rawpedia.pixls.us/How_to_get_Nikon_ICM_profiles

I’m not sure where those files go in WIndows but in my darktable on Linux they end up in my “input color profile.”

I’m tempted to delete that page. Using Nikon ICCs means that instead of starting with a blank canvas you are starting with a pre-drawn coloring book and your options are limited.

It seems to be what he’s looking for though.

If he’s just using it as a starting point what is limiting about it?

Isn’t that also the case with any matrix/LUT ICC/DCP profiles that dt/RT ships with? You need an input profile of some kind, and the more options, the better. Dt has the unbreak input profile module, which modifies the profile-embedded curve (you can e.g. use gamma 1.8 with C1 profiles, etc.). In RT you can only disable the embedded TRC, but then you can build one from scratch using RT’s own curve modules. My opinion, keep that rawpedia page, if only for information purposes – it’s not like it misinforms anyone but can give one ideas about input/camera profiles.

1 Like

So, if I’m understanding what you’re saying here that using one of those Nikon profiles limits me because that input profile uses fewer colors?
And could you explain what the unbreak input profile does again but a little less technical if you could?

David, my previous post was directed to Morgan – I should have marked it more clearly, I suppose.

I don’t think Nikon profiles use smaller gamut than the ones darktable is shipped with. I don’t have the ICC profile dt uses for my camera extracted, but I can show you how a Capture One ICC profile for my camera (Nikon Coolpix A) compares with the one from the NX-D software (Screenshot 1) and one that I made with DCamProf (Screenshot 2).

As you can see in this 2D representation, the NX-D profile is larger in green and red hues (the a axis of Lab), but smaller in the blue/magentas (the b axis) than the C1 profile. My DCamProf profile contains both of them quite easily (except for some of the lightest and darkest tones).

When it comes to the “Unbreak input profile” module: some ICC camera profiles may have an embedded tone curve which is different than the “typical” gamma 2.2. See the manual, and the UFRaw website for some more info on it.

I had asked earlier why Morgan thought using the Nikon profile was limiting so I figured your post was a response to him. But your post did get me thinking and resulted in my follow up questions. Thanks for the answers! I’ve got to look into the DCamProf process.

If you’re interested in DCamProf, see here. Another place would be the Lumariver manual, which is possibly more approachable.

Incidentally, it occurred to me that the NX-D profile that I’ve got is based on the AdobeRGB profile (which was probably set as the program’s working space). I must set NX-D to ProPhotoRGB and then extract that profile to compare.

1 Like

darktable produces pretty good jpegs by default. For indoor pictures I usually correct exposure, white balance and apply denoise stack. For outdoors you usually deal with clipped highlights.
After you have done it a dozen (few dozens :slight_smile: ) times the process goes fast.