EOS Rebel T7 - Linux Mint 21.2 Darktable 4.8.0 - RAW too dark and weird colors while importing in Darktable

Hello I’ve got this situation here, see the image.

A. Linux mint’s Xviewer can see view my RAW .CR2 and .JPG files, both images are exactly the same.
B. I can take the RAW and convert it to TIF with XNview, the TIF will look identical as the .CR2 and JPG.
C. The CR2 and JPG are identical to the result I can see on the LCD of the EOS Rebel T7.
D. When I open the CR2 in Darktable, 0 - Original , i get a green underexposed image.

My question: How can i open my .CR2 in darktable, without making any modification, for it to look exactly the same as the .CR2, .JPG or .TIF viewed from the Linux Xviewer or the camera LCD?

0 - Original

11 - Orientation (11 effects applied automatically is still terrible)

You cannot:
The camera does quite a bit of processing to create the jpeg image. Details of that processing are kept secret by Canon (same goes for any other camera brand). A (smaller) version of that jpeg is embedded in the metadata of the raw file, and that jpeg image is what’s shown/written in your A, B, and C (all show the camera-generated jpeg image)

So using a raw developer like darktable means you have to process the raw image, ideally to a final image you like. Normally, when you open an image in darktable’s darkroom view, a few modules are activated. But if you select the 0 line in the history, you’ll get the absolute minimal processing. That causes the green tint.

Well, that depends on how you look at it. The first five are absolutely necessary to transform the raw data in a displayable image, and perform ‘effects’ that most other editors hide under the carpet (though most do give you some choice in demosaicing). The next 6 steps perform actions you need to get a decent image.
Have a look at the darktable manual to get an idea on how to use darktable.

Hello and welcome to the community!

When you open the raw file, xnview is showing you the embedder jpeg, that’s why it looks the same.

Are you sure its processing the actual raw file? If it looks exactly the same, its probably extracting the embedded jpeg and converting it.

You’re still seeing the jpeg.

You need a handful of modules applied to get a usable image. If you go back to the “original zero” step in darktable, then that is what you see, a green image. That is your actual raw data without much processing to it.

You’re probably not going to get that with darktable. It isn’t built to mimic the jpeg because we don’t know what processing went into getting the jpeg.

You could try and calibrate with darktable-chart but I don’t think that was ever a great option.

Outside of darktable, ART or RawTherapee can use a DCP which might get you a bit closer.

As for darktable, reading the manual might help clear some of your questions up: darktable 4.6 user manual - Overview

I think there’s a basecurve preset to mimic that camera’s processing. That said, the more modern way of processing with filmic tends to give better results (for me at least), but it does require some practice.

The manual should tell you how to select the workflow using the basecurve (but I strongly suggest you use the scene-referred workflow!)

Yeah those are very generalized tho.

Note that most open-source raw developers work the same way. For example, read:


Thanks for the fast replies.

I don’t know if I want to spend the time reading the manual now… I am facing the challenge that my task is to rebuild an image as good as the one created by the camera , which has been created by canon algorithms, perfected over decade of experience. If we merge all the effects of DT together, there is probably over 1 million combinations possible to set an image. It’s mathematically difficult to believe.

Question: After reading the manual, how long did it take you to be able to rebuild your images at the same grade as the Canon secret algorithms, and even better? So I can have a general idea of the amount of time required to be successful.

I get the RAW but it’s It’s like getting a file without a password.

Observation: I tried to compare the RAW and the JPG created by the Camera in Xviewer and honestly they are both very high quality, i need to zoom quite far in to see a bit more noise in the .CR2, viewed from the Xviewer. Like you say, It looks like the CR2 is using an embedded inferior JPG image that Xviewer can read. The original JPG copy of the camera is maybe a bit better. I felt like the CR2 image was way better from Xviewer.

Thanks for that info… it’s not exactly filling me with joy but at least I know the truth.
Left is the CR2, right is the JPG

QUESTION #2: I am not going to buy a Adobe product I don’t use Windows in the first place. But i’ll ask anyway, can lightroom decrypt canon algorithms and give us the RAW directly as we see the original JPG from the camera? or it’s the same story as with DT and RAWTherapee?


I tried to set the basecurve preset with Eos Like setting, but it don’t work, the image is still very, very bad.

Maybe improved a bit but mostly a disaster. :grin:

It depends a bit on the camera - some are pretty easy. Just set dt to the sigmoid workflow, tweak exposure and contrast and you’re 80% there. Others… :flushed:
This is talking generally - but the Canons I’ve used aren’t much different to anything else.

RawTherapee has an automatch tone curve option which can do quite well… although it can’t match it perfectly it’s often close.

Make sure you set color preservation to ‘off’ if you’re aiming for a match.

If you don’t want to read the manual (well, the Overview section at least) or a tutorial, then you will never get anywhere. DT is an advanced raw editor for power users, basically, and doesn’t try to hide that fact. It gives you supreme control, but you need to put in the time.

If you want to faithfully recreate the camera JPEG, shoot JPEG and don’t bother with raw. Even Lightroom, despite Adobe’s best efforts, will not give an identical result. That said, RawTherapee does have an option to try to match the JPEG that might work for you.

Don’t remember, to be honest. But it ultimately depends on your aptitude and how much you practice. But as long as you don’t try to match the Canon look, and instead simply try to get something you’re happy with, it won’t take long.

As already said, you did not actually compare the raw with the JPEG. You compared the HQ camera JPEG with the (somewhat LQ) camera JPEG preview that’s embedded in the raw file.

Simply adding base curve on top of the mosaiced raw data is not the way to go.

Here’s the tutorial part of the manual: darktable 4.6 user manual - an introduction to darktable's workflow

Here’s a video version:

He explains some details that you won’t get by only reading the manual. It’s for an older version, so some things have changed, but when in doubt refer to the manual.

Thanks for this link,

Eek! My Raw Photo Looks Different than the Camera JPEG
→ it’s setting the record straight as rvietor and paperdigits already did.

Currently I have a couple hours of experience using Darktable and almost nothing on RAWTherapee, I installed it to see if the RAWs were a bit better and spent maybe an hour trying it.

NOTE: I understand why they called it RAWTherapee, it’s like some kind of anger management workout through working with RAWs (while trying to believe this is NOT BS :rofl:)

Should I spend time on Darktable or RAWtherapee first to improve my situation? This wiki is about RT. I would pick the one giving the best results faster.

Wow thanks for all of these infos.

I’ll read the manual as suggested.

1 Like

That’s a hard question tbh - it’s very much (IMO) down to personal preference. I prefer darktable as once one’s over the initial hurdle it’s very slick in terms of workflow (again, for me YMMV)

I wrote this little article a while back - couple of things have changed since (I added a comment at the end a bit later as update) but the basics are there. It may or may not help but some have apparently found it helpful.

Also, if you’d like to post and licence a RAW file in a Play Raw thread I (and many forum members) would be happy to show examples of how we’d do it. Which again, may or may not me helpful but worth a try! :)_

I’m not sure why your screen shot shows the raw looking so weird. I just downloaded a sample RAW for the T7 from dpreview.com and it looks pretty decent right off the bat, opening it with dt set to the sigmoid workflow in preferences.

Try applying this .xmp to one of your files and see how it looks. This is just a quick 5min attempt at matching the jpeg with bread and butter modules. (edit - you’ll need to reset the color calibration module after applying it or your WB will be wrong.

IMG_1043.CR2.xmp (8.1 KB)

Referring to this?

He disabled all modules, giving him the actual raw data with only photosite color added. No white balance, no demosaic, nothing at all.

1 Like

Oh, I missed that. Yes the second screenshot looks alright just a bit low on contrast/chroma (compared to the JPEG I mean!)

Thanks. :slight_smile:

While that’s not a bad starting point, the aim is to do “better” than the camera.(“Better” as in “being closer to the image you want to create”.

The camera’s processing is aimed at the taste of the majority of users, and to handle most common situations. But your taste may differ, and there are enough situations where it’s possible to do better than the camera. Once you start mastering raw development (with whatever program you decide to use), you can tailor the result to the image and to your taste.

And of course, one thing the camera cannot do is treat different parts of the image in a different way (dodging, burning, selective sharpening, removal of dust spots, removal of phone lines, etc.) as the camera cannot mask off parts of the image.

I totally need Darktable… I spent hours trying to find the best settings for sunsets and daylight shots with the T7 but I see some limitations.

Too much exposure: The sky is white and the greens turn yellow.
Too little exposure: The sky is nice but the trees are all dark.

Seems like there is no way around post-processing to correct the exposition in scenes with a lot overexposed and underexposed elements.

I’ll wrap up the content of this thread now. People gave me couple of paths to follow including, the most important things to adjust in priority.