I enjoy working with my RAW files and I acknowledge that my camera does a good job making JPG. I want to combine both. Proposed feature : create a “preset” that emulates the embedded JPG. This would obviously not be a preset, but some kind of algorithm that makes the draft RAW development match the embedded JPG as closely as possible for exposure and colors, then adds the DXO corrections and real presets, and then lets the user tweak the RAW file further as he wishes.
Hello and welcome Joachim.
It’s a great idea and I would believe that quite many would love the idea and many other applications - especially within video editing does this as a basic functionality.
Don’t forget to vote on your own proposal!
Yes, a basic thing.
This is, more or less, what the default color rendering does
Color Rendering / Generic Renderings / DxO Camera profile (xxx)
in combination with presets
“1- DxO natural style” or “2- DxO Standard” without SmartLighting.
More or less indeed. Built-in JPEGs are rendered with a default or user-selected picture style which can change the looks in easily visible ways.
The two presets you mention minus smart lighting correspond very roughly to what Canon styles would call “Landscape” and “Standard”.
The following can be done to mimic camera styles more closely:
Get Adobe’s free DNG converter and copy the .dcp files that correspond to your camera to a folder that can easily be accessed by PhotoLab. Then, the rendering can be set to one of the corresponding styles.
Note that this has to be done manually, DPL can’t select a .dcp profile automatically.
I also doubt that DxO will come up with something that automatically selects a third party profile anyway, even though this would be nicely user-friendly.
Err, no, that’s not my experience with both .CR2 files (Canon 400D) and .CR3 files (Canon 90D). Most notably, the DxO Standard preset is much more contrasty and the greens are much more yellow compared to the in camera JPEG.
Thanks for pointing to the workaround. It would be so much easier to have a tool that matches the embedded JPEG by creating settings ad hoc, would it not ? I did as you told, but a) where do I find the Adobe .dcp files ? and b) Adobe DNG result is very neutral, like DXO default setting, not like the embedded JPG. This is a screenshot of two Windows Photo displays : on the left, the NEF file (the embedded JPG, I suppose same vivid color as what DXO displays for a fraction of a second before switching to RAW preview); on the right, the DNG file)
But from where would PL get the JPEG settings? They may or may not be baked into the file, they don’t always exist in the EXIF data and if they do, they are in a manufacturer specific EXIF section that may or may not contain relevant and translatable settings.
I address that question in the proposal. The proposal is not to apply real or supposed camera JPG settings, but to create a tool, an algorithm that finds the settings by comparing the RAW development preview to the embedded JPG file.
That is going to be complicated. Different colour spaces and profiles come into the mix and how do you cater for those images that need to be adjusted locally to get a similar result? How do you cope with images that are under-exposed but that could be recovered?
JPEG files have a much smaller dynamic range and pixel depth than RAW files, how would you cope with that?
PhotoLab is not a magic wand, it is a toolbox that requires knowledge and discernment to get the best out of.
Looks simple on the surface, but:
the embedded jpeg might have some corrections applied, like distortion, crop, NR,
but the information on how the jpeg was produced is in proprietary format.
jpeg is a lossy compression with its own (micro)artifacts.
jpeg is only 8-bit deep (it’s not jpeg2000), while RAWs are typically 12, 14, or 16-bit.
there would be problems identifying the colorspaces used/intended.
Sounds like a big project.
I am not suggesting that this feature should get the best out of the photo or of dxo. I am suggesting that this “emulate embedded JPG” would be an optional starting point for further editing, just like any other preset. Except that it’s not a preset, but a tool. And not a perfect copy of the JPG, but a good approximation. Much like other tools for auto-exposure, etc.
technically it can be done, with fine print about optics correction, by creating on the fly a camera profile ( or LUT at least ) …
you have raw data and you have embedded OOC JPG in raw file - this by itself is a shot of target where the scene is a target and OOC JPG gives data for target “patches” ( as I noted - optics correction shall be addressed )…
downscale both to the same pixel dimensions, pixelate/grantulate/posterize next (to reduce the amount of “patches” / colors to some more sane value ) and then build a lut or straight camera profile for this image and apply… for advanced users provide options to control lut grid size, lut smoothness, etc …
this is what commercial tools like 3DLutCreator or some free lut creating tools do in some way… camera profile apps like dcamprof ( free ) or its paid GUI version LrPD can be used too with additional tools… but a raw converter itself can simply do this on the fly
PS: now optics corrections - that can put things out of alignment - so while raw converter like DxO PL can execute the process internally w/o applying optics correction a fine print shall be made that raw shot shall have camera optics correction turned off ( granted some cases like with extreme optics corrections for Canon RF16 or whatever you can’t turn it off in camera - but for most regular lenses you can )
PS2: and of course that limits conversion to the color space of OOC JPG, but AdobeRGB is good enough for many folks, even sRGB for that matter
PS3: easier way is to finally make profile creation inside DxO PL a proper one - create LUT profiles, from arbitrary targets, and give people an option to enter target data with “use OOC JPG color values” for a target as an option… that again is what tools like 3DLutCreate, dcamprof / lrPrD, etc do - just you have it inside raw converter… a convenience feature for those who want a quick way to approx OOC JPG rendering
While not exactly what you are suggesting, I created my own PL OOC jpeg preset. It gives me a very close match to the OOC jpegs produced by the camera.
Note, I have adjusted the OOC jpeg camera settings to my preferences and the PL preset is based on that, not the manufacturers’
The PL preset is based on a dcp profile and tweaked to match the OOC jpeg for colour and overall tone. It also reduces the reds slightly as I find the reds in the OOC jpegs to be a bit bright - this aspect is sometimes desirable though.
I can then apply other corrections if needed.
However, sometimes that preset just does not work and I use other presets that I created.
For me, the current DxO Optical Corrections Only preset already does much of what is proposed. Given everything that is (or should be) on DxO’s plate now, this does not seem a priority ask. Not pooh-poohing the idea, though.
OP is talking about easy way to replicate off the bat stuff like OOC JPG colors, brightness / contrast / tone curve
Yes, an optimization solving a luxury problem. One way to avoid the problem would be to NOT briefly display the embedded JPG when opening the photo file for the first time. That would reduce or eliminate the wish to keep the profile of that preview.
I think ON1 PhotoRAW has such a feature called AI Match. It’s supposed to automatically “set the Tone & Color sliders to closely match the embedded JPEG preview in your RAW file”. I don’t know if that works, I never tried it.
It would be a great feature, especially to emulate Fuji film recipes. However, to my ears it sounds like a lot of AI guesswork and I can’t imagine the results would be consistently good.
not everything that is labeled as “AI” is “AI” , if it does whatever any plain LUT generator does calculating luts from 2 side by side pictures there is no “AI” in it needed
free LUT generators like
DxO FilmPack provides a load of presets, a bunch of which are Fuji “digital film” emulations.