[WorBry]

I am looking at the remaining three clips btw but I have to say they are quite challenging, especially using only the Channel Mixer to make the color corrections.

In the 'Cave' clip (cave-GOPR0238) the B&W striped Sergeant Major fish make an appearance again, but if I use them for neutral reference I'm fighting marked yellow discoloration of the 'skylight' opening and color variegation in the aura around it.

Best strategy looks to be:

1. Apply a 'soft clip' to pull-down that super-white data with a smooth roll-off gradient - easy to do in Resolve, maybe could be done in VDub2 with (gradation) curves, I'll have to see. Incidentally, your clips are all recorded with full range luma (0-255) and all of the grading in Resolve is being done at full 'data levels', so there's no forced hard clipping (limiting) going on. Obviously in VDub2 the full range flagging on decode ensures that 'PC.709' conversion coefficients are applied by default.

2. Use the 'skylight' opening as neutral (daylight balanced) reference for the channel mix in the first instance.

3. Adjust the mix to achieve best compromise with the 'white' stripes on the Sergeant Major fish.

I'm making headway but hitting that fine balance is finicky to say the least. I'll post once I've arrived at something presentable. I'm determined not to resort to qualifiers and the like, as I want to replicate the process in VDub2.

BTW, did you use a dive light in that clip? Looks like the Sergeant Major fish closest to the camera are receiving some artificial illumination. If so, might explain why I'm finding it so hard to hit a good compromise.

As for the other two clips. Difficulty in both is finding an identifiable neutral reference to go by. GOPR0201 at least has some skylight so I could try to apply the above strategy. There's also that shoal of really tiny fish that are possibly black and white - maybe can sample off those at really high magnification.

But GOPR0233, nothing, not even sure what color the fish there are supposed to be - bluish grey maybe ? Only inroad I can see there is to start 'rebuilding' the Red channel with an initial RG/RB sub-channel shift, align the Red and Green channel profiles as closely as possible (referencing Histogram), see what colors that looks to be pulling-up and plumb for a color mix that is in someway believable.