ACES Look Transforms

Tags: #<Tag:0x00007febd06c03e0>

(Andy) #1

Hello,dear ACES friend

The jpeg has a red box--Show look 3DLUT – Look transforms-- LMT(show look),Please check more from

We want to learn more how to convert the 3DLUT to the LMT.

Looking forwards to your profession guidances

Thank you so much


(Jim Houston) #2

There is a June thread that has good info on the LMT…

In short, LMTs are a Linear-In to Linear-Out conversion to an ACES color space file.

A general 3DLUT from another place can be converted if you know what the colorspace_in and _out
are, and what transfer function and quantization are being used in that LUT. But that process can be tricky.

If the LUT has the final look baked into it, then you would want to use a path that takes the output
of that LUT for a particular device and invert it through an InvODT/InvRRT, to get rid of both the RRT and ODT tone curves. This will limit the dynamic range of the file to whatever device the 3DLUT was made for.

To go further, I would have to look into the .ari format a bit more, so perhaps someone else can jump in.


(Daniele Siragusano) #3


I do not understand the point in going “forward Print Film Emulation (PFE)” -> “Inverse ODT Inverse RRT” -> “RRT+ODT”, from a systematical and educational standpoint.

This reads the same as “apply the FilmLut and then enter ACES through the ‘bottle neck back door or Rec.1886/709’”

LMTs constructed this way will undermine one of the biggest goals of ACES which is short- and medium term “future-proof-ness”. If you apply the LMT and back it in your ACES Linear AP0 archive master you basically archive rec709 or P3 but not the “indirect output referred” - scene referred data.

Also, your LMT will have all the caveats a PFE brings when applying to digital data, like out of gamut errors etc.

Going inverse ODT/RRT might look ok when again viewed forward RRT/ODT, but it can produce highly unstable ACES Scene referred data. If you apply gradings after the LMT you can reveal errors which were hidden before. So combining different LMTs to produce new LMTs is not really possible or highly unstable.

It is much harder to construct good LMTs which are not limiting the dynamic range and gamut and at the same time produce the same intended look. Also if the “Display Rendering Transform” introduces too much “look” by its own it is harder to produce another “look”.

I hope some of this makes sense

(Daniele Siragusano) #4

About the Arri Document:
I think their definition of LMT is a bit wider, which could go from a simple CDL grade, to a more sophisticated grade all the way to a “complex Look”.