Input vs Output LUTS.

Hi I was wondering if anyone could clear up what the difference between a 3D input Lookup table is and an Output lookup table. I have a project shot in Arri log and noticed they both seem to do and render out the same. Is there a particular time you should use one over the other? Thanks in advance!

Hi I was wondering if anyone could clear up what the difference between a 3D input Lookup table is and an Output lookup table. I have a project shot in Arri log and noticed they both seem to do and render out the same. Is there a particular time you should use one over the other? Thanks in advance!

So input LUT is upstream of your grade, output LUT is downstream. Let's say you are grading an arri shot and the LUT you're using is arri to rec.709. If you're apply the input LUT your image is now 'baked', ie if the LUT clips anything — it's clipped and your color correction nodes will not see that info that was gone because of LUT application.
If you use the output LUT that is arii to rec.709you will still see the rec.709 result but you will have an access to an entire recorded image information while you grade, because transform doesn't happen until the last step in the image processing pipeline.
Please someone correct me if I am wrong.

Hi I was wondering if anyone could clear up what the difference between a 3D input Lookup table is and an Output lookup table. I have a project shot in Arri log and noticed they both seem to do and render out the same. Is there a particular time you should use one over the other? Thanks in advance!

So input LUT is upstream of your grade, output LUT is downstream. Let's say you are grading an arri shot and the LUT you're using is arri to rec.709. If you're apply the input LUT your image is now 'baked', ie if the LUT clips anything — it's clipped and your color correction nodes will not see that info that was gone because of LUT application.
If you use the output LUT that is arii to rec.709you will still see the rec.709 result but you will have an access to an entire recorded image information while you grade, because transform doesn't happen until the last step in the image processing pipeline.
Please someone correct me if I am wrong.

Thanks Dimitry, so to me it sounds like an output is a better way to go since its non-destructive, so to speak. What might be a situation where you'd use input over output? Seems to me that output is the most beneficial.

To illustrate, here is an image with gain turned down and an input transform LUT arri to rec709 applied as a global setting:

With an input LUT

input00086400.jpg (294.17 KiB) Viewed 8075 times

And here is the same correction but with lut applied as an output LUT, also as a global setting:

Output LUT

output00086400.jpg (294.82 KiB) Viewed 8075 times

See how different the highlights are. In the case of the input lut there is no way to bring them back. Lowering the exposure just makes them a uniform gray area as opposed a uniform white area.

As a somewhat helpful analogy I would say an input LUT in this case makes footage look like it was shot with a standard rec.709 camera. And then you grade.
In case of the output LUT it is if it still being shot with a rec.709 HD cam for the final delivery, but when you grade you change the light before it gets to this virtual camera. Difference being that unlike changing the light of the real world scene, you do so as interpreted by alexa and its color space.

If you would like the same effect on a clip by clip basis, you may right click on any node and apply a LUT there.
Input LUT option there is that lets say you have to quickly process a bunch of files, or if you might have a custom look LUT that was used for onset monitoring and the creatives would like to have dailys on a non-LUTtable devices with the same look.
Output lut is there to serve many other purposes, like conforming to a specific projection device, or as a more flexible option for getting into a ballpark for rec.709 deliverables in the case of the particular LUT we are discussing.
One can get the image to look good without the use of LUT, it really just depends on your workflow.