The DT setup - I do a few things which are slightly weird, like hardlight blend a coarse generated grain into the picture to try and break up the low-frequency chroma noise, color blend the exposure adjustment, linearize the base curve (because I’m manipulating the output gamma with so many blended adjustments it’s excessive), etc P8264809.ORF.xmp (9.7 KB)
For gimp, the main moves are using the 32-bit float linear format (gimp 2.9 gegl 0.3), and levels + channel mixers instead of curves (which are too coarse in adjusment and mix up the colors), and lots of blending layers. With the way gimp works, either I’m blending copies of the imported image, or various composites of the work I’ve already done (using “New From Visible”). Didn’t save the working file from last night (the challenge was just to rip something off in a few minutes) so I’m doing a super-fast workthrough again to give you an idea of how I would work through something like this. I’ll write the layers down step-wise (bottom to top), in reverse order from the way they work in gimp (top to bottom):
“Normal” (the base, background layer)
“Linear Light” (Levels: B=0 G=4.8 W=100, BO=3.5 WO=90)
“Hardlight” (Levels: B=9.8 G=4.75 W=41, BO=11 WO=52)
Sometimes I’ll do tricks like a “vibrancy” type move, where the top layer is “New From Visible” (rendering a composite and putting it on the top of the stack), blending as LCH Color and using levels to get an effect (colorfulness of bright tones vs dark tones, overall colorbalance etc can be adjusted like this). Normally I wouldn’t use such an aggressive blending mode as LinearLight because most images are sufficiently contrasty enough (even as a scan off of film) that Softlight or Hardlight work well to target certain ranges of tones or color. In this case trying to pick up the faint stars and get better differentiation from the background needed super big moves.