@charlotte yes i can see the plushie ❤️

@eri @charlotte as someone who’s way too knees deep into computer graphics, they’re not “just” uncompressed images, they’re images that you can’t send directly to a framebuffer, they’re images with values that don’t make any coherent sense, they’re images that the pipelines used to modify and show them to the screen weren’t GPU-accelerateable until a decade ago, they’re images where you can forget about colourspaces altogether because that sometimes just isn’t embedded into the image, they’re images that CG people on the PBR and realism side spent decades trying to replicate the processes of rendering of because it’s hard as shit even when you can control all the parameters and have a perfectly clean image. they’re images whose research laid the foundation for RGB gamer HDR monitors and any lighting techniques more advanced than the blinn-phong model. an uncompressed image would be a .gif or a .bmp not a raw

for a software like GIMP, there’s a ton of assumptions about images made from the very beginning of CG history that only apply if your image is linear RGB with premultiplied alpha in the integer range 0-255. there’s a ton of assumptions about how colours are meant to be blended and how they’re meant to be rendered. even for Krita that has experimental support for colourspaces that’s not just different Srgb profiles, it’s slow and clunky and doesn’t work with most of the tools you’ve come to expect to work out of the box with RGB and unoptimised as hell because no hardware support, on top of having to translate the entire image to Srgb anyway for displaying because no hardware support. and alpha blending probably doesn’t work as you’d expect either because premultiplied alpha is an assumption that can’t be applied to a lot of colourspaces because it assumes that both the colour components and the alpha component are linear so you can linearly blend the colours without unnecessary multiplications and divisions

all of that in the framework of “there may not be gpu acceleration and there might never be” and having to write both the software and gpu pipeline. and then the reality of “handling raw images only being possible to be gpu accelerated since a decade ago”.

and then there’s the part of “yes you can bake the RAW image into something you can use with the rest of the pipeline” but like, it’s extremely nontrivial to implement. there’s a ton more stuff about RAW that’s hard that I’m not mentioning because I don’t work on RAW image editors, but i know that so much of the things that i do look at and work on are dependent on the research into how to make digital cameras from “impossible wet dream” to “almost trivial enough to be widespread”

@twinkle @eri yeah that is what i tried to racconvey with https://akko.chir.rs/objects/a387d45a-7e1c-4c98-9d2a-e6d696b7eeae

It’s as racclose as i could turn a raw into a png with darktable while disabling as much processing as possible

like sure the general Shape of the image is accurate, but color is missing, lighting is wrong, the relative brightness differences are Wrong (because it’s linear) and the file is also more massive than the original