Lumteri Color Panel Stacking miltiple Lumetris

Thank you very much, I am reading this twice to find the best stacking solution for my needs! 🙂

Ok ... bonafides. I work for/with/teach pro colorists for pay, right? I've been around far more colorists than editors over the last decade plus, and spend nearly as much time in Resolve as I do in Premiere. I've discussed color issues with most of the leading team members of Premiere at NAB for the last decade plus also.

Where to start ... well, first thing I'll go with is you have to understand processing order no matter the app used. In Premiere, that's the clip itself, starting with the top effect in the ECP going down. Within Lumetri, it's top down as is the order they appear in the Lumetri panel.

After the effects directly place on the clip are processed, then processing moves to the lowest Adjustment layer.

And within that layer, is again, top down in the ECP. 

If you have an additional AL above that one, it is processed after the lower AL.

And again, within the that layer, processing is top down in the ECP.

So ... applying a log->linear conversion LUT in an AL above a clip, then doing work on the clip itself, is ... working pre-LUT. Which you clearly are not expecting, and that is where you are getting one of your mystery things from.

Though for a mathematically correct log->lin LUT based workflow, trimming the clip pre-LUT is something that does need doing. And, as you are not correctly doing your log->lin step anyway, that's the next subject.

Proper Log->Lin Workflow Methods

You can do this either of two ways ... the old school and Ancient Order of the Past method, which is what you kinda sorta mush through now. Using a LUT. 

Or you can use the modern and preferred method, quicker and far safer for your pixels, and use the built-in algorthms of Premiere or Resolve to do the log->lin process with mathematical calculations, NOT with a simple chart of this-to-that stuff, which is all a LUT is.

LUTs are called "the dumbest math out there" by colorists for a reason. There actually isn't any math in any LUT ... a Look-Up Table ... or LUT ... is simply a text chart, quite literally. With triplicate sets that state that X triplicate (RGB data) goes to y. Such as that 23/08/32 becomes something else. All determined by numbers in that table, and in-between, the app guesses, trying to get the correct linear progression mapped out.

And while LUTs can be very useful, if you don't know 1) exactly what any LUT does to the image, 2) the proper image data for that LUT to work as expected, and 3) how to trim the image into the LUT, well ... you can have all sorts of problems. 

Applying a log->lin normalization LUT without being able to trim pre-LUT is bad. If the LUT causes any pixels to go above 100IRE, they're clipped, gone ... and you can't get them back post-LUT. Any pixels gone below 0, again, gone, and you cannot get them back. Anything over-saturated is simply blocked color, and there isn't a fix.

And understand, I've learned how to create and test LUTs for years. I've learned how to deconstruct what LUTs are doing, to roll my own for use instead. In Premiere, Resolve, SpeedGrade, and LutCalc. I've years of heavy use of LUTs in my kit. That said?

The newer, mathematical algorithms are mostly fantastic, as they will normally never allow for crushing blacks, clipping whites, or over-saturating some channel. All of which are normally expected with most LUTs when applied to field-produced media. (Which is why you have to trim the clips pre-LUT, of course.) 

And they are faster to use. Have your cake and eat it too ... truly! But if you do need to use a LUT for log->lin normalization, here's how.

Correct Manual Log->Lin Normalization with a LUT Method

I would only recommend this under either of two conditions. First, you've got a client that demands X LUT be used on their stuff. At that point, oh ... well. Second, your camera is not producing a LUT that Premiere has an algorithm for. Again, at that point ... oh well, use the manual LUT method below.

You need to apply the LUT, then trim the clip to fit through the LUT, remember? The best way to do that in Premiere is:

1.  Apply the log->lin normalization LUT in the Creative tab of Lumetri, applied on the clip itself.
2.  Now go to the Basic tab of that Lumetri instance, and use the Exposure, Contrast, Sat, and other controls to trim the  image correctly "through" that LUT. Watching both the image and your scopes. (This is simply to get a tonally useful, basically neutral view of the image.)
3. Use the other controls in the tabs of Lumetri, and in additional Lumetri instances, to finish the grade.

The Safer And Faster Method For Log-Lin Normalization 

This uses the mathematically superior algorithms available for many current camera log forms. It is by definition nearly incapable of mangling pixels via clipping, crushing, or over-saturation. So no "trim" work pre-algo is needed ... as all  your pixel data is still available after the algo.

1.  Setup your color management options correctly in Premiere, now ALL found on the Lumetri panel's Settings Tab ... the one named Settings.
2.  That would be setting auto detect log and auto tonemapping to on.
3. Sequence (normally) to Rec.709.
4. DONE.

No Lumetri, no grading work at all. Slick and swift.

Ok ... will it look a bit different than after your LUT? Of course ... because all log->lin normalizations involve both technical and aesthetic choices in the making. There isn't any "perfect and only!!!" normalization for any media made by any camera. Including any produced by the maker. Which is why Sony and many other makers give the users several different LUT based normalization options.

So an algo based log-lin normalization, like any LUT based normalization, is simply used to get the most neutral, starting image possible. And if you want to have a specific mod to all your media from one camera, that's easy. 

Do a basic grade of the things you always want to see in your clips. On the clip, in one Lumetri instance if possible. Use two if you must.

Now in the ECP, select your Lumetri grade, whether one or more instances, and "Save Preset".

This now gives you a saved Preset of that "look" in your Effects bin, and can be quickly drag/dropped over a whole bin of selected clips to apply it to all selected clips at once.  Very fast and efficient.

It's called then a "Source Effect" in Premiere, and will be applied to any instance of any of those clips anywhere in the project. And can at any time be modified if any clip needs it ... simply go to the Source tab in the Lumetri panel, and you are working on the Source level application of Lumetri.

And spiffy, this is applied after the auto-normalization via the algo.

You may have heard of Walter Volpatto? One of the "biggest" names in colorists? A point he stresses all the time is that too many people do far more complicated stuff than what actually is needed, and will show on the viewer's screen no matter how it's delivered. His typically is BIG screen delivery, but that doesn't matter, it's the same whether by screen, streaming, broadcast, DVD/BluRay, or YouTube.

Coloristst are taught some basic Truths. Until you learn the following, it's incredibly frustrating.

Every screen out there is very, very different from yours. 

And no one, ever!!! ... will see exactly what you see while grading. It is physically impossible.

I bet this article here helps out, it's from Neat Video the denoise company. It explains the order of processing.

https://www.neatvideo.com/blog/post/pr-order