Color Correcting With Blending Modes – Recovering Highlights With Multiply

I am a great fan of using blending modes, but I am not sure about what you are doing when you show how the “highlight control” doesn’t work I assume you are talking of the shadow/midtones/highlight log control and you use them with the control surface so we cannot see it?

Fascinating insight. Gonna try it on my next job. My previous go-to method was just like you showed, making a new node as my first adjustment and luma keying the highlights, then either gain down or highlights control down, sometimes gamma down. But yeah, it always tags the nice highlights in people’s faces! Also curious how this multiply method compares to setting up a node at the start of your tree that’s simply gain down in log mode with a dialed in high range.

Good look at using blending modes in Resolve … will try that this week. I’ve done a fair amount of work in PrPro/Lumetri, and *had* to use Lumetri on the clip plus another layer on an adjustment layer and use either multiply or softlight to recover details or at least make the rolloff to “blown” more palatable. But I’ve not thought of trying blending in Resolve … with it’s rather wider palette of tools.

For my understanding of blending modes, multiply just multiplies one pixel by another one, but that kind of mathematical operation should be used in a linear environnement. I assume Resolve is managing the gamma/linear transform internally for us to see just the gamma corrected/video screen friendly picture, but the operation is probably applied in a linear color space.

Anyway, for what I understand the multiply blending mode is helping here to get an image where the highlights are almost “isolated” – and in a smoothly manner – from the rest of the tonal range. Think of this : in linear, all the pixels values are between 0 and 1. Multiplying two values under zero always gives a lower value that’s why you get a darker image. Now when you multiplying highlights by themselves you get a lower that still not too far away from the original value (0.9*0.9=0.81). Now when you’re multiplying lower values by themselves you get lower values (0.2*0.2=0.04), the lower the lower, increasing in a logarithmic curve. And from this “almost smoothly isolated highlights” image you can pull a nice key and recover your highligts.

Is that make any sense?

This was a life saver.
Thanks, Dan!