I struggle to reproduce out of camera jpegs with Darktable

Hi,

I really struggle to obtain edited raws looking the same as in camera jpegs using Darktable.

I am using a sony a6400.

Here is an example where I tried to recreate the in camera jpeg by editing the raw. But I am unable to make it look the same.

Which modules and settings would you use to make the edited raw look the same as the in camera jpeg?



I don’t even try to make my edits look like the camera jpegs… (for one, I find the Sony jpegs too contrasted, and often too sharp and saturated). And to be honest, if your main aim is to get camera jpegs, why bother with raw?

But I mainly use “exposure” (usually +1EV as a starting point), “color calibration”, “filmic”, “color balance RGB” (preset “standard”), and one or two “diffuse or sharpen” presets. Those modules are either activated automatically, or gathered in a style. I may also use “tone equalizer”, “denoise (profiled)” and “lens correction”. The last two are just switched on, and don’t require further adjustment (in most cases).

“Color calibration” (white balancing) is probably the most critical step to get your colours “right”, and not all that easy when there’s no reference in the image. Also because when there is no reference, it’s not clear what the right colours are…

3 Likes

In darktable 5, you may find a style for your camera, but it won’t match perfectly.

On the lighttable, select some images, and:

You can select styles in the darkroom, too, and there’s even a Lua script to apply them automatically:

Also added a Lua script to auto-apply the appropriate style on import and manually apply styles to a collection of previously-imported images.
(darktable 5.0.0 released | darktable)

Start here: darktable user manual - process
And be sure to read darktable user manual - process

What you need to remember is that the camera’s rendition is not ‘the truth’, it is one possible rendition, which you may or may not like. Some commercial vendors sign NDAs with camera manufacturers to learn what processing the camera performs; with open-source, that’s just impossible.

If you really want exactly what the camera does, stay with the JPG, or go with the manufacturer’s own software, or with one of the ‘big guns’.

If you want to make the photo look like you want it (as opposed to how the manufacturer envisioned a satisfying image), stay with darktable or one of the other open-source software that allow you to tweak every aspect. BTW, RawTherapee and ART support DCP camera profiles that come with in-camera curves; you may also give them a try to get a more ‘like the camera’ starting point.

1 Like

Hi, thank you for your answer! I try to reproduce out of camera jpegs with darktable as a way to improve my editing skills. I think it is great to understand how the camera processes the raw files.

1 Like

Hi, I try to reproduce out of camera jpegs with darktable as a way to improve my editing skills. I think it is great to understand how the camera processes the raw files.

The styles seem really nice! But can we see the modules used to obtain the styles or is it just like a look up table? Because I would prefer the former to really understand the applied processings.

The styles apply a set of modules with preconfigured settings. Apply the style, and then check what modules it added (image history stack in the darkroom, left-hand column, right under the navigation thumbnail).
https://darktable-org.github.io/dtdocs/en/module-reference/utility-modules/darkroom/history-stack/
https://darktable-org.github.io/dtdocs/en/darkroom/pixelpipe/history-stack/

ok that’s really great thanks!

Each camera manufacturer invests time and resources to produce JPGs to make their cameras look good to the consumer. Sure, if that is what you want then use the camera JPG and forget about RAW. However, opening the JPG in DT and then taking a snap shot gives you a bar to meet in your editing and this can allow you to make your own style. I did this for my Canon R7 and it gives me a great starting point but ultimately I can usually produce a better result than the cameras JPG.

First I start by working out what exposure is needed to match the look of the cameras JPG. Dt has a default of +0.7 EV but I found with my Canon R7 that +1.1 EV was closer. I also use sigmoid and not filmic so I discovered that sigmoid defaults with a contrast setting of 1.7 worked better for my Canon R7. I then decided that appling local contrast module at default values was good as well as the shadow and highlights module at default values. And while the as shot white balance in the color calibration module was good for the neutral tones I found I had to go into the color xzones module and increase the chroma of some of the the zones to better match the camera’s JPG. I then applied denoise (profiled) module at defaults and some initial sharpening. This was all done while comparing the snapshot from the JPG. The result was not a prefect match but it was close and the raw file had a sharper more detailed look with similar exposure, color and contrast. Or follow @kofa advice and use one of the included styles that match our camera. That is what inspired me to create my own style.

I’m a Sony photographer too, and I actually like my camera’s (a7iv) jpegs and am quite happy to use them when I don’t have time to post-process. Although, before I went over to the raw side, I used to do some tweaks with GIMP.

I do my initial culling with jpegs. A hard cull, where if I delete the jpegs, a script then deletes the associated raws before I import into dt. So I know what the jpegs look like, but they are not my criteria for the finished image. But if dt starts with something like the jpeg, then that, for me, is a great start.

For me, darktable-with-sigmoid gives me a good start, and that’s largely the reason that I stuck with dt. I wasn’t that keen on the new camera styles, but only tried a couple of pics.

That suits me. But I doubt if it is a one-size-fits-all answer for everybody. Quite probably it depends on what and how we photograph, and on what results we want.

I’m not sure you can learn that from trying to imitate the camera. Nor am I convinced it’s all that useful: cameras have to process images fast (or they won’t be able to reach those 10 images/s they like to advertise with). But that means they have to use simple and fast algorithms, not necessarily the best, for any task (e.g. sharpening or denoising).

I’d suggest you read the section about workflow in the manual (that’s about 4 pages), then work on a few of your images where the jpeg looks good. Use the jpeg as a rough guide so you’ll know when you mess up things (then just clear the history stack and start again, I still do that regularly :wink: ).

There’s nothing wrong with using the provided styles and presets (that’s why they are there). But I’m afraid if you get too used to using prebaked styles, you won’t understand what darktable does and can do, so when you hit an image that’s harder, you have no idea how to tackle that image.

That’s certain: I’m dealing with some microphotos at the moment, and they require “slightly” different processing than e.g. landscapes or flower close-ups… The in-camera jpegs are not much help either…

4 Likes

I tried to compare the SOOC technically with the edited raw but the images are different sizes, so all I can say is that the one “looks like” the other.

2 things I cans think about:

  • colors are off so you have several options ranging from :
    • color balance
    • color calibration (my favorite) with the addition of channel mixer
    • color balance rgb (hue shift plus 4way tab)
    • primaries
    • …
  • second the contrast and black level ares slightly off so you can :
    • cheating with the not so recommended ‘haze removal’ ← fast and dirty fix
    • adjusting output with your tone mapper of choice (see black level and contrast in sigmoid for example) ← really easy as well
    • use the 4ways tab of color balance rgb) ← my fav for this application

I can of like this kind of exercise but unfortunately I do not have the time right now to have a try at your example.

I know this is your stated goal but IMO (which isn’t worth much) this might be your biggest waste of time.

For sure you can use the jpg as a sort of general reference but really learning to use the tools to craft the raw data to your vision or a better or different version is the power of having and using a raw capture… You have the camera jpg strait away…and it might be bang on or crap either way you have that already. If its good then simply use it…

Much of this as noted above will depend on what you are shooting and what you intend to do with the final results.

For me two of the most powerful things that are important to focus on are composition and lighting… the first you can do in jpg or raw but the second one is much better done in raw. You can use this to enhance the subject you pick for your photo. In the playraw section of the forum some of the most impactful edits come from a simple crop choice to alter compostion and the lighting choices and often they look nothing or very little like the jpg …

So time spent learning to get the tools to move your image in certain ways is the key (DT can be a bit of a challenge that way as there are often multiple paths to the same destination). Then defining your subject and composition and finally lighting it to set the mood or to direct the eye or to give it color. Maybe its just me but chasing a jpg match doesn’t free you up to explore and master that crutial part of editing…again maybe its just me…

For sure I started out thinking that way and gauged what I was doing against the jpg from the camera but it was only when I abandoned that and took the approach of understanding color and digital editing concepts that I could make much better use of the tools.

If you spend an hour pushing and pulling sliders and tools to try to make your image as sharp as contrasted etc etc as your jpg you might get there but in the end those edits are going to work for a small subset of simlar images and when applied to another one they might just make a mess.

Given a new very different photo might require a very different approach and as you modify certain elements and see how the data reacts you might be moved to take the image in a certain direction that is really awesome and yet again not at all a match to the jpg…

So use them as a guide but cut the cord and dive into the tools and follow the playraw entries and you will get much more out of it…again IMO…

2 Likes

If you are not limited to use DT only and willing to explore other OpenSource programs then I’d recommend to take a look at RawTherapee and ART. With Auto Matched Curve or Adobe DCP profiles you can easily get pretty close colors. I spent a lot of time and effort learning DT but still find that it is usually quite difficult for me to get colors and contrast similar to OOCJ. I used Nikon Z6 and now Sony A7IV.

1 Like

Would you be willing to submit a photo in the “play raw” category so you can see how we may approach developing it?

1 Like

Has anyone remembered that the so-called “fine” OOC jpeg is normally only 4:2:2 chroma sub-sampling and surely one would want 4:4:4 (no sub-sampling) when editing a raw, i.e. better than the embedded or the OOC jpeg?

[edit]by which I’m agreeing with earlier questions “why try to match the OOC JPEG”.

Seems like the JPEG quality is not a factor for @BorisRabit [/edit]

Hi, I will maybe do it!

2 Likes

I am not sure I understand your query. If you are getting the desired output from the in-camera jpeg, why would you want to achieve the same using DT?

On the other hand if you want some other style of output, then using DT makes sense. I for one try to create a film like output where I am able cover the entire dynamic range. This I achieve through, primarily, three (maybe four) modules - Sigmoid, Tone equalizer, Color Balance RGB and color calibration

Hi, thank you very much :slight_smile:

I want to do it as an exercise :slight_smile: