sorry for if this question has been answered, I’ve been looking for it but didn’t find a concise answer …
I am trying to improve my postprocessing and I have probably very basic question or set of questions regarding the precision and color profile. I use mainly RawTherapee + GIMP for to develop the RAW photos from my cameras (Nikons D850 and D7200 mainly) … I’ve been always using the defaults without thinking much about them
RT uses by default working profile “ProPhoto” which is I guess OK and output profile is RTv4_sRGB (and other options are _Rec2020, _Wide, _Large) and 16bit integer uncompressed TIFF … I use these files then in GIMP and I ofc keep the color profile on importing in GIMP so I effectively work in GIMP with 16bit integer/perceptual gamma (sRGB) precision …
Q1.1: Should I use 32bit tiffs ? I mean while exporting from RT ? And does floating point Vs integer really matter ? Will I gain some sensible improvement in IQ when working in 32bit (float) Vs 16bit (int) ?
Q1.2: Should I use Linear lights or Perceptual gamma (sRGB) ? Latter is GIMP’s default on any tiff imported into him … I’ve recently seem some David Davies video on youtube elaborating on this point a bit and came into impression that Linear Light is better and I should use it as my default.
I’ve been investigating a lot on the difference between the quality of edits made in RT on RAW Vs doing that in GIMP over the 16bit/32bit tiff …
Q2.1: Can I use tiff to for instance brighten (add exposure / shadows / highlights) the image without penalty to IQ Vs doing that in RT ? Actually what are the data flexibility when exporting the image from RT with 16 or 32 bits (float Vs integer) in tiff for edits like exposure (tone balance) and colors Vs doing same in RT over the RAW data ?
Q2.3: is there any difference in duplicating Vs creating new visible layer ? I mean if I will open 250MBs bit tiff it will have one layer … Is there a difference in the new layer data when I will create new layer by duplicating existing layer Vs creating new layer from Visible ?
Rationale:
I am trying to improve my postprocessing workflow to maintain max possible IQ and to just optimize it regarding the tools I use … I was using RT (or DT) and GIMP many years without really thinking in the defaults … I tend to use RT for global edits and then I export the image and continue processing that in the GIMP and I do local edits there. I typically try to prepare my picture in RT as much possible so the changes in GIMP are rather subtle typically … I also use a lot of exposure blending or focus stacks blending … I do it all in gimp with layers after exporting tiff files from RT …
I will go slightly off on a tangent in my answer so please forgive me. When I used RT I would export as 16 bit tiffs and use layer and masks in gimp to combine two or more edits from RT. This was often simply a darker sky in one edit and a brighter foreground in another edit of a landscape image. These were then combined as layers using a mask created with the gradient tool.
However, 99% of my GIMP work disappeared when I switched to DT because I could do localised adjustments including gradient masks to combined two or more exposures in a single edit. However, if I do decide to do further editing in GIMP, DT allows me to export as an xcf file with 32 bit floating point for further editing in GIMP. I see 32 bit floating point as an advantage in GIMP, but it may be overkill and 16 bit is certainly more bits than the original raw file.
I don’t feel qualified to say much about color space, but it may be best to keep a large color space as long as possible in the editing process. But I defer to more knowledgeable people on this.
However, 99% of my GIMP work disappeared when I switched to DT because I could do localised adjustments including gradient masks to combined two or more exposures in a single edit.
That’s also my experience, and I’m not yet using masks in DT
As to 32-bit FP… Gimp works internally in 32-bit linear, so if the image is 32-bit linear once loaded in Gimp, you skip a lot of format conversions so performance is better, but you need more memory. However for the kind of work we do on photos this isn’t too much of a problem.
I don’t like on DT that concept that you create a mask per tool and that you create mask and do more edits within that …
can somebody please help me to understand this:
Q2.1: Can I use tiff to for instance brighten (add exposure / shadows / highlights) the image without penalty to IQ Vs doing that in RT ? Actually what are the data flexibility when exporting the image from RT with 16 or 32 bits (float Vs integer) to tiff for edits like exposure (tone balance) and colors Vs doing same in RT over the RAW data ?
Q2.3: is there any difference in duplicating Vs creating new visible layer ? I mean if I will open 250MBs bit tiff it will have one layer … Is there a difference in the new layer data when I will create new layer by duplicating existing layer Vs creating new layer from Visible ?
I think there is no difference between duplicating and creating new visible layer in pixel precision. However if original layer has mask, creating new visible layer does not copy the mask.
Additionally I’ve found, that if I will export image into the jpeg from 16/32bit floating point / linear light precision in GIMP to jpeg a lot of programs will have a problem with floating point and will screw colors … so just avoiding that fancy Linear Light in GIMP at all
I think it is the problem of using linear light in low bit depth (Jpeg) image. It reduces tonal gradation in mid and shadow range than sRGB. And GIMP uses linear image in internal image process, so I don’t think there is any need to specify explicitly linear light color profile.
