The G’MIC project has a neat grain synthesis option using a grain sample as a seed. That might be worth investigating?
Thanks I will check it out 
Sorry, I didn’t mean to be dismissive if I came across that way. For me, everything usually passes through GIMP for something before I’m done and I have used grain overlays for a long time in that fashion.
I agree that a grain module would be nice (and I’m positive the devs do as well), but it’s really a matter of getting competent programmers and maintainers on-board the project to implement those things. Unfortunately, those types of folks are in short supply - so things are prioritized differently.
It’s ok Thanks, I am actually a Junior C++/C# developer, but I don’t have much experience. I want to check out the sources just for my personal interest, maybe play with the issue a little bit (it’s distributed as an opensource sw under gpl if I am not mistaken).
Thanks for the info about the topic.
(I don’t want generally participate to hi-jacks but will still answer this one)
No. This new behavior has been introduced for a single reason : local editing, where users will have to be able to draw outside of the sheet. Yes, local editing is still planned, but delayed due to GUI problem. The core engine will also have to evolute strongly before anything good can be done on this front.
I don’t want to open new discussion for the topic, so using this old thread…
I was really missing grain on my photos, but don’t like Darktable way of doing it because it is removing details and the feel is somehow artificial and unsharp for me (+ I prefer using RawTherapee).
I’ve found workaround which works for me. I’m shooting on higher iso (800 - 6400, preferably closer to 6400) with my Fuji X100s. Demosaic it with 1-pass Markesteijn, set super high sharpening level, set super high local contrast, set contrast by detail higher for finest details, lower for coarsest details and lower the contrast in the exposure tab. Using that usually in combination with Ilford HP5 film simulation.
The photo is now containing sharp “grain” which is adding some randomness to the picture, so it don’t look so “sterile”, but still keeping details on the photo.
Some of you is using similar trick? Please, share your settings.
There isn’t any film grain modules, but use capture sharpening with lots of iterations and zero contrast threshold for some pleasant fine grained results, amplifying the noise that is already there in a decent and filmic like way.
Instead of using lots of unsharp mask, try using RL deconvolution or Capture sharpening in the raw tab with zero contrast threshold, and lots of iterations. Also, use the wavelet edge sharpening module. This will prevent oversharpening of the edges, while pleasingly increasing fine texture and inherent grain that is already present.
Wavelet tool looks a bit too magical for my taste. But Capture sharpening + higher local contrast seems to be good starting point.
Hi @maaca,
Even if this information is somewhat old by now, it might well be of good use to you:
Have fun!
Claes in Lund, Sweden
1 Like
Contrast by Detail Level does just a few things the Wavelet Levels tool does, so they are similar in concept, but wavelets gives you more power, such as control over lower frequency detail (big macro-gradients), residual image, and boost cut relative to the amplitude of the detail.
It also allows manipulation of the detail level chroma, and wavelet sharpening that can target either fine detail or edges, and more.
Well, I would be more comfortable reading «Contrast by Detail Level does just a few things the Wavelet Levels tool does». It’s more accurate.
Moreover, CbDL is applied before in the pipeline, while WL is applied near the end, and that offer some advantages over CbDL.
And last, but not least, one of the biggest benefits of using WL is being able to edit the residual image, not just the details in each level.
Wavelets are not magical. They are just complex, and our main problem is that the documentation hasn’t been updated yet (in English), but the Denoise and Refine module, in addition to the Residual Image module are an awesome help for any image. I would even say that Capture Sharpening and those 2 modules are the biggest advantage RT has compared to other software.
@maaca, as @Waveluke said, you can use Capture Sharpening and RL Deconvolution (controling haloing) to enhance noise. And in Wavelet Levels, it’s better using the lower sliders of the Denoise and Refine module, specially in levels 1 and 2. Maybe even in level 3.
1 Like
The thing is that I understand how are the wavelets working, in theory. But I’m missing documentation which would help me to use that in real world.
My camera (the demosaic algorithm?) is generating noise which I like. I just want to make it more distinctive.
When I was experimenting with wavelets, I was able to generate new/different patterns, but somehow the tool didn’t help me to keep and increase the existing ones.
We worked together, a few months ago, @XavAL and me, so that he wrote educational documentation for wavelet
This documentation was written in Spanish and needs to be translated into English (my very low levels in these 2 languages obviously do not allow me to do it).
Xavier started his work by reading the initial documentation which was in French and wrote by me (and partially translated in English).
During this exchange, I was able to explain concretely to him how the wavelets actually work.
And I think that between the beginning of our discussions and the end, he understood that wavelet and its language are different from the usual concepts (luminance, contrast, etc.)
But different does not necessarily mean complex. Complexity is something relative it depends on your level of knowledge and practice.
In conclusion i call users not to flee wavelets because they fear complexity, but to discover its use, which makes it a remarkable tool (both in the “main” version and the “local” version - branch newlocallab)
jacques
2 Likes
I’ve spent a bit more time with the wavelet tool. I still think that it is magical tool, but now in the positive point of that word. First results are looking very good and are closer to my idea than what I have achieved with “normal” tools. But still, it would be great to have a bit more complete documentation for that.
