#apertus IRC Channel Logs

vup: you mean pixels?

I have profiled several film stocks

then you profile your camera and your intended film stock with extensive profiling before principal photography

then when you are on-set you have your LUT all ready to go for monitoring, and then in post you hopefully use the same development procedure and do the per-shot stuff after.

another note is that this all gets significantly harder in the new HDR landscape. if you are profiling a film stock, for example, you need to know in advance what displays you are outputting to, because that will dictate the maximum values you need to actually account for when profiling the LUT in meatspace

and then of course, you have all the spatially-varying artifacts to consider like grain, halation, and a host of other things. there's already a ton of commercial products competing in this space with fully-fledged products because the second a colorist gets a little technical the first thing they think of is to start emulating film, haha

to me the key ingredients right now are still:

- ascertain the native gamut

- - "good" varies subjectively but i like cineon and am used to grading in cineon-like spaces. many colorists seem to agree here

regarding our customers, realistically we are targeting spunky indie filmmakers right now in narrative. i can't comment on other industries, but i know that high-end features work at such scale, with so much money flying around, that an open source newbie is just completely not on the radar. our big use cases for a long while will be individuals and small teams on small budgets who are technically curious and want to do something a little different.

btw i'm not an authority in any sense on the marketing stuff, just spitballing the way i see it. i'm sure others who have more history in the project can comment more on that stuff

ex: film is interesting because it is a complex system with simple inputs and controllable but dynamic outputs

because of course if you just profile film stocks, you are in essence just copying and pasting inputs/outputs of some already-established system rather than deciding what is important in a more fundamental way

i think it's good, because color scientists have long worked to perfect it. also, we are just used to it that films look this way

yeah, i mean the default argument for emulating film is "it has been engineered over decades to look good" so that's fair enough. i just think not enough people are thinking about _why_ they think stuff looks good and just kind of copy what film does as an exercise for lack of better ideas

RED's log space is nice to work in, I have used it before. good middle gray and decisions about the primaries in wide gamut

thanks

i do think yedlin is a great resource too, though, to be fair, i just like throwing a little shade sometimes

his text is the best ressource on that perspective i found :)

btw if you actually want to know the nuts and bolts of making a LUT I have done it a bunch. meaning like how you actually calculate the sparse data linear interpoloation and how to know if your data is "good" or not so i am happy to share my knowledge

just going through a few of the other items on your LUT doc page, some of the generic luts like Phantom LUTs and stuff is really just that, generic. they aren't really turning a camera into another camera in a meaningfully precise way.

but it would be interesting to have a "Film look" LUT for the axiom beta that resembles the ARRI color science

and maybe open source tools to play around with looks

sadly the digital camera world is full of a lot of snake oil salesmen when it comes to luts and "looks" and "film emulation"

i don't mean this to be argumentative, i think we agree. i just think that we need to avoid making ambiguous goals without being really crystal clear on what the intention is, when we invoke a phrase like "film look" or "arri color science"

to me arri is not doing anything we cannot do fundamentally given their sensor architecture, from a color standpoint. they just happen to have good data and are working on sound principles. we can do that too. the only big limitations i see right now are more on the hardware side than on the sensor or anything

yeah, huge differences

for two, the 35 is designed hand-in-hand with the new LogC4 curve which is meant to address a full HDR mastering pipeline

four, the 35 has some experimental new features that let on-set DPs bake in certain things directly to thte RAW file, like noise profiles (which you can think of as "grain") in order to impose some creative control over the DI

this is a great example of arri knowing their market and listening to customer feedback. it's a clever feature that is polarisiing, but opinionated. i like opinionated.

that's right

i mean a LUT has nothing to do with spatially-varying things like gran, which is what the ycan change. or bloom, filtering, things like that

is a polarizing filter the only thing that can not be changed in post? there are even quite qood effects for halation (in resolve), so a mist filter is not needed (but maybe easier to use)

it's just polarizing in the sense that increasingly in digital camera land, more and more control has gone into the realm of DI and post-production

so, less about the tech, it's more like an emphatic statement on arri's part that they support people who want to nail the look in-camera, that want to have some idea of how the final image is at the time on set, at the expense of some flexibility later. but this is not forced on users, it is merely an option. but still a spicy one

i see. thanks