Thanks for the explanation of the additional feature in the Highlight reconstruction module. You are an excellent educator. Your videos fill a need and you have made it easier for me to learn/use DarkTable. I recommend your channel to others whenever I can. I think you have it right when you said that you are going to do photography in the future. From that experience, ideas for new educational content will emerge. If not, go do more photography that inspires you. Your channel is refreshing (e.g., not about gear and personal trips).
9:13 : yes, you are doing something wrong XD Your clipping threshold is set > 1 which should practically disable the module. If the white balance and raw coeffs are detected correctly, threshold = 1 is the max. But then, some cameras have raw coeffs issues (unidentified at this point), such that you may need to use thresholds as low as 0.7. Then, you need to set the reconstruction radius to roughly twice the size of the biggest blown areas. Then, turn up the iterations until the neighbouring colors get inpainted. 12:30 : dt is close to feature-completeness, so that's less shiny new stuff to be excited about. On my side, developing the dt UCS color space has taken like 4 months full-time. The guided laplacian thing took something like 2 months too. As underwhelming as it may be, getting color spaces right is not a luxury and between getting the theory right and testing all the theory in practice, there is a lot more involved than just pissing code…
Thanks for the clarification Aurélien! When I get home from this trip, I'll revisit the module and see if I can make more sense if it. As for the shiny new stuff, totally understood. We all appreciate the fantastic work you do, even if we don't understand it! 😊
Hi Bruce, It is not the frequency of your videos which retains us but the excellent and unbiased presentation. Given the specular highlights on your car's wheel , panels and bumper, I would approach that image with the healing tool in the retouch module. In the wheel a circular selection could be fitted in to relieve the clipped area, On the panel the same, only easier, and the chrome trim strip in the bumper could respond well to a path selection repeated a few times. 2 - 4 minutes work (and the front wheel as well).
The problem with this technology is that there is a feeling that it always needs to be updated. You must get to a point where 'if it aint broke don't fix it'. Adobe have got to adopt this model to keep getting people dissatisfied with what they have so that they can keep up their sales. You don't have to do this with Darktable. It is excellent as it now stands. This is coming from someone who has eventually left the Windows platform for my photographic needs (CS4 & Lightroom 3 - so I do practice what I preach) and migrated fully to Linux for everything I do (having used Linux for the past 15years for everything except photography). P.S. Thanks Bruce for great videos. You have enabled me to transfer from Lightroom in a structured way.
Hello Bruce, delighted with this return and good future holidays. I think that there are quite a few users who think that version 4 is much more consuming in system resources for so-called improvements, good example than collections... I'm waiting for Aurélien's fork under windows which seems to have cleaned up the source code a lot, in the meantime I'm on an intermediate version 3.9 which is still running correctly. Warmly
Another great video, mate. I've read / seen somewhere if you turn filmic off, then reconstruction works better, only after you turn filmic back on and did some more tweaking there, witch seems a lot more work than it should be. Now that darktable is slowing in development, maybe it's time for a new feature in your channel, some sort of 'how to' with your images and complete workflow in darktable, just a suggestion. Keep up the great work.
As for the processing of images, I do a lot of that for the patrons, which is why I don't do it too often on the main channel. But, I'll keep the suggestion in mind.
I think you are right that it will be a per image as to whether this works better or the reconstruction in Filmic. Although I'm not sure why they are doing this when they have the reconstruction in Filmic??? Nice one, enjoy your trip.
It is a two-step process, I first select a reconstruction method in filmic, and if I am not satisfied then I try to wrangle the highlight reconstruction module.
While it was definitely fun having tons of new tools coming out every 6 months, a release with more under the hood stuff happening than lots of changes to how you edit photos is probably nice for newer users to have a program thats not changing as rapidly when they're first trying to learn it.
Thanks for taking the time to make this and put it up. I'm sure that many will have got a great deal out of it - something that applies to all your clips. Sadly, I stopped watching after you read out the sentence starting, "This mode is..." because I am committed to Fujifilm. I seem to recall a time when DT would not open Fujifilm RAW files at all, so I guess there has been some progress in this regard.
Yeah, I'm not 100% sure how disadvantaged Fuji users are by these occasional "X-Trans sensors not supported" caveats. I guess Fuji users find a way to deal with those issues.
@@audio2u it's difficult to say if products like Capture 1 or Lightroom address this and other "Bayer only" features that DT offers. One thing for sure, DT offers extraordinary power for very little outlay, and thanks to your good self and Boris Hajdukovic the DT user base is bound to be expanding. I for one, am exceedingly grateful to you both.
Thanks for the video Bruce! I've struggled with this module and I also use L channel a lot. Using color or the new option does often leave artifacts but the magenta or other color remnants can often be improve by pushing them off the histogram with base curve, exposure or RGB curve and look quite good; I'm not a fan of filmic and don't like it being turned on automatically now. Other times I've used color zones and drawn masks to desaturate the remnants and they look quite good. As you say, it's image dependent but often worth experimenting to get the best results if the image matters. For stuff I don't care highly about, I find L-channel is the least work.
You don't HAVE to have filmic applied by default. Go into preferences and change the setting under the processing tab for default something-or-other.... Can't think what it's called right now, but the two options in the drop down menu are 'modern' and 'legacy'. Switch to legacy.
Hopefully the next update you Darktable gives us something more. I feel the same. And as a Fuji shooter, I get a bit annoyed with what's missing, but recently I also bought an Olympus camera, and can see I'm not actually missing anything exciting for the X-Trans sensor.
You need to ask yourself if darktable is a tool meant to do a job or a new toy to be excited about. At some point, you will have all the tools you need in your shop and what will remain is simply to get the work done. But if you want a cool shiny new toy to feed your GAS, you are on for disappointment. The most important part of the software is the pictures you put in it and the sensibility you put into them, not the clutter around. The clutter around may give you a sense of power, but that's out of place in any artistic endevours. Violinists still play on an instrument that hasn't changed one bit since the 18th century. Seems to do the job alright.
@ I've been on the Darktable Facebook group for years saying that there needs to be a cut down version, cutting out the duplicate tools and improving the performance and reader of use. Darktable needs a HSL tool improvement badly, the Colour Zones module is really badly designed compared to the competition. Improvements in usability in areas like this is all that will make us happy I think. Have a slider for each colour, just make things more accurate. I love Darktable, and it's my main editor, and I appreciate the continual improvements in Filmic and highlight recovery, but there's a lot that can be removed, a lot that can be refined and more that can be added to make it an even better program. I appreciate your work, I really do, and I can see why you have branched your own version. So, you can understand others frustrations with the software.
You are definitely right that this wasn't a big release in terms of important new features. I guess they just got too close to having to do a 3.10. As for the bayer sensors vs. x-trans. I've not really run into many limitations with Darktable for my xt-30. There are a few features that may not work but nothing I seem to need or care about. All the new scene referred stuff works great with it. And as this new feature shows, it's a bit niche and not that impressive to begin with. Highlight reconstruction was pretty good to begin with before this version. There seems to be a bit of feature interaction between this module and flimic which Aurelien Pierre tried to explain in a recent video. Did not quite get all of that but I think the gist of it is that highlight reconstruction happens early in the pipeline. Hence the sensor type matters probably for the new laplacian mode.
Glad to hear that the sensor in the Fuji doesn't cause that much concern! As you say, it's fairly niche stuff, and the HR has always worked really well anyway. I too watched Aurélien's video and felt that he wandered a little off topic, but I didn't want to accuse him of that in case it was just ME not understanding the connection... Which is more than likely the case! 😊
I did a video of laplacians module and as you, I find LCh and Reconstruct colour better for most of my images. Now, between LCh and Reconstruct, it is really image to image. Sometimes I fond one is better than the other. I always try both. Trying Laplacians, results are best when the area is tiny. A blown sunset is not a good example. I also tried up to 16 iterations with the largest diameter 2048 (very very slow), it took 20 seconds to compute. I don't know what slow and fast are... I only have a Macbook Pro 2020 (M1), with 16GB RAM. One of my viewer tried the same on his PC and it took 2 min 21 s... I was also having the idea to move to Fuji. I see that in fact Fuji shaked the status quo of the Bayer Filter. Camera manufacturer is using it, but it is the best? So they have a sensor nobody uses and it takes some times to build compatible software and filters... Open source will be even slower to develop software for that sensor. That doesn't surprise me... About darktable 4, I was waiting for big change. darktable jumped to a major version (3 to 4), and not minor. In the Open source software community, this is big... But it feels a bit flat for a release, I agree. darktable 4 should have been named 3.10.
Fuji X-Trans need different maths than Bayer sensors, it's not just about catching up from the software side. Demosaicing methods designed for Bayer can't be used for X-Trans, and the other way around, so it's worse than duplicating code, it's actually different logics entirely. But the fact that Fuji medium format cameras use Bayer seems like an acknowledgment that X-Trans sucks. The whole idea of the X-Trans design was to have more pleasing sensor noise…
hi Bruce and thanks for your description about highlights reconstruction in this version 4.0.. but why clipped RGB channels are not just "White" ?? ..my old lightroom 5 is more simple and work better on all my sunsets.. darktable is maybe more sophisticated but a challenge most of the time.. anyway, i'm always happy to learn more details with you!!
Why are RGB channels not just white ? Well, because your pipeline doesn't know what white is before white balancing. But white balancing (the accurate one, the chromatic adaptation) needs a fully-formed RGB pixel, meaning it needs to come after demosaicing. But demosaicing clipped channels leads to fringing around the clipped areas. So we need to recover clipped areas before demosaicing where there is no white balancing and no color profiling, meaning we don't even have colors, we have "signal" that voids any attempt of desaturating because saturation is the distance to achromatic and we don't know what achromatic is early in the pipe. Why does LR5 work better ? Well, because it does an inaccurate white balance before demosaicing and therefore you have some rough definition of white there. But you will pay that shitty white balance with LED lights, noticeably the infamous blue lights that every concert stage has now. Just see how Lightroom has saturated blue gradients taking a shift into magenta… So, yeah, you got your white alright, at the expense of everything else… People having opinions on stuff they don't nearly understand, episode 245…
@ yes i understand the pipeline but when RGB channels are clipped (clipped mean 255/255/255 on RGB channels, 0 to 255 in 8 bits format), nothing can be more white than this.. i never saw an other photo process program with this kind of darktable interpretation.. so in darktable, what is the white level for to see a true white?? ..because if i over expose more in darktable, i only have more magenta.. so, to have a true white become a challenge each time that the sun is in my pictures.. this should be simple to deal with, but is not the case in darktable.. my opinion is the pipeline may be a science point of view but for to process photo, it's too much a challenge.. take few sunsets as examples, i would like to see how easy you deal with (a suggestion)..
_Only works with Bayer sensors_ Considering Fujifilm's small market share and the X-Trans-Sensor's market share who couldn't blame software devs (especially FOSS devs with no money doing it in their free time) for going for the bigger platform. As Linux user I think you know exactly what I talk about when I write about (not) creating software considering the size of the target audience.
Hi, I think that the magenta color is caused by filmic and its v6 chrominance preserving mostly. When using v5 everything gets better for my camera (Olympus OMD).
2 ปีที่แล้ว +1
@@jeseterprod No it's not. Filmic preserves what colors are there in the first place, it doesn't invent new ones. Garbage in, garbage out. You need to fix that magenta before the pixels enter filmic.
The new segmentation method is very nice... I have tried the Laplacian a few times with really no success. I think given the new method I won't ever use it...
2 ปีที่แล้ว +1
The segmentation method inpaints flat colors, doesn't recover gradients.
If you're lacking inspiration for videos perhaps I could send you 2 or 3 Fuji XT2 raws and you could do a few start to end work flows on Darktable 4. I'm new to Fuji and Darktable so would be cool to see how a Darktable expert can work through the most useful modules and see what Fuji files trip up Darktable.
I do, but I'm not going to post that publicly either! Can you find me on one of my social channels? They are all listed in reach video... Insta, Twitter, Facebook, etc
Hi Bruce, it is better to have less video's with meaningful content then regular video's with no news or repeated content. If You feel you have something meaningful to explaine do it like you have done before, because that has been great. What I would like to see added to Darktable is a module to manually add / edit EXIF data to a file. With new or vintage manual lenses you do'nt have electric contacts on the lens -> no EXIFdata exchange. So manually adding lens data would be great, don't You think?
![[Darktable] Broken 'Pink' highlights and how to fix them](/img/n.gif)
Thanks for the explanation of the additional feature in the Highlight reconstruction module. You are an excellent educator. Your videos fill a need and you have made it easier for me to learn/use DarkTable. I recommend your channel to others whenever I can. I think you have it right when you said that you are going to do photography in the future. From that experience, ideas for new educational content will emerge. If not, go do more photography that inspires you. Your channel is refreshing (e.g., not about gear and personal trips).
Thanks for the kind words, Alan. :)
9:13 : yes, you are doing something wrong XD Your clipping threshold is set > 1 which should practically disable the module. If the white balance and raw coeffs are detected correctly, threshold = 1 is the max. But then, some cameras have raw coeffs issues (unidentified at this point), such that you may need to use thresholds as low as 0.7. Then, you need to set the reconstruction radius to roughly twice the size of the biggest blown areas. Then, turn up the iterations until the neighbouring colors get inpainted.
12:30 : dt is close to feature-completeness, so that's less shiny new stuff to be excited about. On my side, developing the dt UCS color space has taken like 4 months full-time. The guided laplacian thing took something like 2 months too. As underwhelming as it may be, getting color spaces right is not a luxury and between getting the theory right and testing all the theory in practice, there is a lot more involved than just pissing code…
Thanks for the clarification Aurélien! When I get home from this trip, I'll revisit the module and see if I can make more sense if it.
As for the shiny new stuff, totally understood. We all appreciate the fantastic work you do, even if we don't understand it! 😊
Hi Bruce, It is not the frequency of your videos which retains us but the excellent and unbiased presentation.
Given the specular highlights on your car's wheel , panels and bumper, I would approach that image with the healing tool in the retouch module. In the wheel a circular selection could be fitted in to relieve the clipped area, On the panel the same, only easier, and the chrome trim strip in the bumper could respond well to a path selection repeated a few times. 2 - 4 minutes work (and the front wheel as well).
Yes, using the retouch is a great alternative, given the poor results from the Highlight Reconstruction module.
Thanks for the kind words Andrew. Yeah, retouch would work for sure.
The problem with this technology is that there is a feeling that it always needs to be updated. You must get to a point where 'if it aint broke don't fix it'. Adobe have got to adopt this model to keep getting people dissatisfied with what they have so that they can keep up their sales. You don't have to do this with Darktable. It is excellent as it now stands. This is coming from someone who has eventually left the Windows platform for my photographic needs (CS4 & Lightroom 3 - so I do practice what I preach) and migrated fully to Linux for everything I do (having used Linux for the past 15years for everything except photography). P.S. Thanks Bruce for great videos. You have enabled me to transfer from Lightroom in a structured way.
Thanks for the kind words. Glad you were able to ditch windoze completely! 😃
Hello Bruce,
delighted with this return and good future holidays. I think that there are quite a few users who think that version 4 is much more consuming in system resources for so-called improvements, good example than collections...
I'm waiting for Aurélien's fork under windows which seems to have cleaned up the source code a lot, in the meantime I'm on an intermediate version 3.9 which is still running correctly.
Warmly
Fair enough! 😊
Another great video, mate. I've read / seen somewhere if you turn filmic off, then reconstruction works better, only after you turn filmic back on and did some more tweaking there, witch seems a lot more work than it should be.
Now that darktable is slowing in development, maybe it's time for a new feature in your channel, some sort of 'how to' with your images and complete workflow in darktable, just a suggestion.
Keep up the great work.
Ah, interesting. When I return from the roadtrip, remind me of that, will you? The whole "disable filmic, do HR, then re-enable filmic" thing.
As for the processing of images, I do a lot of that for the patrons, which is why I don't do it too often on the main channel.
But, I'll keep the suggestion in mind.
Thank you for the video!
No problem! 😊
I think you are right that it will be a per image as to whether this works better or the reconstruction in Filmic. Although I'm not sure why they are doing this when they have the reconstruction in Filmic??? Nice one, enjoy your trip.
It is a two-step process, I first select a reconstruction method in filmic, and if I am not satisfied then I try to wrangle the highlight reconstruction module.
While it was definitely fun having tons of new tools coming out every 6 months, a release with more under the hood stuff happening than lots of changes to how you edit photos is probably nice for newer users to have a program thats not changing as rapidly when they're first trying to learn it.
Good point!
Thanks for taking the time to make this and put it up. I'm sure that many will have got a great deal out of it - something that applies to all your clips. Sadly, I stopped watching after you read out the sentence starting, "This mode is..." because I am committed to Fujifilm. I seem to recall a time when DT would not open Fujifilm RAW files at all, so I guess there has been some progress in this regard.
Yeah, I'm not 100% sure how disadvantaged Fuji users are by these occasional "X-Trans sensors not supported" caveats. I guess Fuji users find a way to deal with those issues.
@@audio2u it's difficult to say if products like Capture 1 or Lightroom address this and other "Bayer only" features that DT offers. One thing for sure, DT offers extraordinary power for very little outlay, and thanks to your good self and Boris Hajdukovic the DT user base is bound to be expanding. I for one, am exceedingly grateful to you both.
Thanks for the kind words! :)
Thanks for the video Bruce! I've struggled with this module and I also use L channel a lot. Using color or the new option does often leave artifacts but the magenta or other color remnants can often be improve by pushing them off the histogram with base curve, exposure or RGB curve and look quite good; I'm not a fan of filmic and don't like it being turned on automatically now. Other times I've used color zones and drawn masks to desaturate the remnants and they look quite good. As you say, it's image dependent but often worth experimenting to get the best results if the image matters. For stuff I don't care highly about, I find L-channel is the least work.
You don't HAVE to have filmic applied by default. Go into preferences and change the setting under the processing tab for default something-or-other.... Can't think what it's called right now, but the two options in the drop down menu are 'modern' and 'legacy'. Switch to legacy.
hi, for me with .fit files the highlight reconstruction function mentions non applicable, why?
I'm sorry, but I don't know what an .fit file is.
@@audio2u thanks for the reply, I think they are the raw ones, they come from the ZWO ASI astronomical camera, after staking with the Siril program.
Wow. Completely new to me. I have no answers, sorry.
Hopefully the next update you Darktable gives us something more. I feel the same.
And as a Fuji shooter, I get a bit annoyed with what's missing, but recently I also bought an Olympus camera, and can see I'm not actually missing anything exciting for the X-Trans sensor.
Good to hear!
You need to ask yourself if darktable is a tool meant to do a job or a new toy to be excited about. At some point, you will have all the tools you need in your shop and what will remain is simply to get the work done. But if you want a cool shiny new toy to feed your GAS, you are on for disappointment. The most important part of the software is the pictures you put in it and the sensibility you put into them, not the clutter around. The clutter around may give you a sense of power, but that's out of place in any artistic endevours. Violinists still play on an instrument that hasn't changed one bit since the 18th century. Seems to do the job alright.
@ I've been on the Darktable Facebook group for years saying that there needs to be a cut down version, cutting out the duplicate tools and improving the performance and reader of use.
Darktable needs a HSL tool improvement badly, the Colour Zones module is really badly designed compared to the competition. Improvements in usability in areas like this is all that will make us happy I think. Have a slider for each colour, just make things more accurate.
I love Darktable, and it's my main editor, and I appreciate the continual improvements in Filmic and highlight recovery, but there's a lot that can be removed, a lot that can be refined and more that can be added to make it an even better program.
I appreciate your work, I really do, and I can see why you have branched your own version. So, you can understand others frustrations with the software.
Great video as usual. Could you revisit the image and try the reconstruction with filmic rgb module's highlight reconstruction to compare?
Thanks.
If I remember when I get home from this roadtrip!
You are definitely right that this wasn't a big release in terms of important new features. I guess they just got too close to having to do a 3.10.
As for the bayer sensors vs. x-trans. I've not really run into many limitations with Darktable for my xt-30. There are a few features that may not work but nothing I seem to need or care about. All the new scene referred stuff works great with it. And as this new feature shows, it's a bit niche and not that impressive to begin with. Highlight reconstruction was pretty good to begin with before this version.
There seems to be a bit of feature interaction between this module and flimic which Aurelien Pierre tried to explain in a recent video. Did not quite get all of that but I think the gist of it is that highlight reconstruction happens early in the pipeline. Hence the sensor type matters probably for the new laplacian mode.
Glad to hear that the sensor in the Fuji doesn't cause that much concern! As you say, it's fairly niche stuff, and the HR has always worked really well anyway.
I too watched Aurélien's video and felt that he wandered a little off topic, but I didn't want to accuse him of that in case it was just ME not understanding the connection... Which is more than likely the case! 😊
I did a video of laplacians module and as you, I find LCh and Reconstruct colour better for most of my images. Now, between LCh and Reconstruct, it is really image to image. Sometimes I fond one is better than the other. I always try both. Trying Laplacians, results are best when the area is tiny. A blown sunset is not a good example. I also tried up to 16 iterations with the largest diameter 2048 (very very slow), it took 20 seconds to compute. I don't know what slow and fast are... I only have a Macbook Pro 2020 (M1), with 16GB RAM. One of my viewer tried the same on his PC and it took 2 min 21 s...
I was also having the idea to move to Fuji. I see that in fact Fuji shaked the status quo of the Bayer Filter. Camera manufacturer is using it, but it is the best? So they have a sensor nobody uses and it takes some times to build compatible software and filters... Open source will be even slower to develop software for that sensor. That doesn't surprise me...
About darktable 4, I was waiting for big change. darktable jumped to a major version (3 to 4), and not minor. In the Open source software community, this is big... But it feels a bit flat for a release, I agree. darktable 4 should have been named 3.10.
Yep, got to agree on all those points! 😊
Fuji X-Trans need different maths than Bayer sensors, it's not just about catching up from the software side. Demosaicing methods designed for Bayer can't be used for X-Trans, and the other way around, so it's worse than duplicating code, it's actually different logics entirely. But the fact that Fuji medium format cameras use Bayer seems like an acknowledgment that X-Trans sucks. The whole idea of the X-Trans design was to have more pleasing sensor noise…
Interesting! I was not aware of any of that!
enjoy your trip
Having a ball so far!
@@audio2u having a ball i like how you role lol
hi Bruce and thanks for your description about highlights reconstruction in this version 4.0.. but why clipped RGB channels are not just "White" ?? ..my old lightroom 5 is more simple and work better on all my sunsets.. darktable is maybe more sophisticated but a challenge most of the time.. anyway, i'm always happy to learn more details with you!!
Great question. Wish I knew the answer! 😊
Why are RGB channels not just white ? Well, because your pipeline doesn't know what white is before white balancing. But white balancing (the accurate one, the chromatic adaptation) needs a fully-formed RGB pixel, meaning it needs to come after demosaicing. But demosaicing clipped channels leads to fringing around the clipped areas. So we need to recover clipped areas before demosaicing where there is no white balancing and no color profiling, meaning we don't even have colors, we have "signal" that voids any attempt of desaturating because saturation is the distance to achromatic and we don't know what achromatic is early in the pipe.
Why does LR5 work better ? Well, because it does an inaccurate white balance before demosaicing and therefore you have some rough definition of white there. But you will pay that shitty white balance with LED lights, noticeably the infamous blue lights that every concert stage has now. Just see how Lightroom has saturated blue gradients taking a shift into magenta… So, yeah, you got your white alright, at the expense of everything else…
People having opinions on stuff they don't nearly understand, episode 245…
Thanks again, Aurélien!
@ yes i understand the pipeline but when RGB channels are clipped (clipped mean 255/255/255 on RGB channels, 0 to 255 in 8 bits format), nothing can be more white than this.. i never saw an other photo process program with this kind of darktable interpretation.. so in darktable, what is the white level for to see a true white?? ..because if i over expose more in darktable, i only have more magenta.. so, to have a true white become a challenge each time that the sun is in my pictures.. this should be simple to deal with, but is not the case in darktable.. my opinion is the pipeline may be a science point of view but for to process photo, it's too much a challenge.. take few sunsets as examples, i would like to see how easy you deal with (a suggestion)..
_Only works with Bayer sensors_
Considering Fujifilm's small market share and the X-Trans-Sensor's market share who couldn't blame software devs (especially FOSS devs with no money doing it in their free time) for going for the bigger platform. As Linux user I think you know exactly what I talk about when I write about (not) creating software considering the size of the target audience.
Yep, fair point!
Hi Bruce, great video. is the purple coloration due to the reconstruction in FilmicRGB?
Not entirely sure, to be honest!
Check out this video - He explains some of the pink / purple coloration.
th-cam.com/video/z-zVMHKTBBY/w-d-xo.html&ab_channel=StudioPetrikas
Hi, I think that the magenta color is caused by filmic and its v6 chrominance preserving mostly. When using v5 everything gets better for my camera (Olympus OMD).
@@jeseterprod No it's not. Filmic preserves what colors are there in the first place, it doesn't invent new ones. Garbage in, garbage out. You need to fix that magenta before the pixels enter filmic.
@ Dear Aurélien, I agree, but it is very difficult to me to deal with this magenta *before* the filmic module. Do you have any suggestion?
The new segmentation method is very nice... I have tried the Laplacian a few times with really no success. I think given the new method I won't ever use it...
The segmentation method inpaints flat colors, doesn't recover gradients.
Yeah, as I said in the video, I tend to find reconstruct in lch works best in most instances.
If you're lacking inspiration for videos perhaps I could send you 2 or 3 Fuji XT2 raws and you could do a few start to end work flows on Darktable 4. I'm new to Fuji and Darktable so would be cool to see how a Darktable expert can work through the most useful modules and see what Fuji files trip up Darktable.
BG, sure, upload a RAW to a file share of some description and flick me the link.
@@audio2u I don’t seem to be able to post a link here. Do you have email?
I do, but I'm not going to post that publicly either! Can you find me on one of my social channels? They are all listed in reach video... Insta, Twitter, Facebook, etc
@@audio2u Hey Bruce, try your studio email. Thanks.
Hi Bruce, it is better to have less video's with meaningful content then regular video's with no news or repeated content. If You feel you have something meaningful to explaine do it like you have done before, because that has been great. What I would like to see added to Darktable is a module to manually add / edit EXIF data to a file. With new or vintage manual lenses you do'nt have electric contacts on the lens -> no EXIFdata exchange. So manually adding lens data would be great, don't You think?
Man, that would be awesome! But I'm not sure whether that is a simple task. I suspect it's not.
darktable 4.0 is the same as Kernel 6.0 - not really what you would call a _major update_
Right!
All options do ashit job