Lacq’ing in Depth

If you’ve used Avid editing systems, you’re probably familiar with the “film dissolve” transition effect that is offered alongside regular video dissolve. Film dissolve is supposed to morely closely approximate the look of a traditional optical dissolve effect created by printing film frames on top of each other at varying exposures.

I don’t claim to know the algorithm actually used by Avid, and it’s been a while since I’ve used their systems… But logically something similar should be achieved by blending the images in linear light colorspace, as done for the glow in this previous post.

Because Conduit can be used as a transition filter in FCP, it’s easy to roll our own dissolve effect. Let’s try it with these two still images first:

Here’s a frame from the middle of the dissolve. The first is the video dissolve — just a Blend node at 0.5 bias. The second is the same thing with linear light conversions applied:

The difference is fairly clear: in the video version, the station image is obviously cut with black, and the hands are much more pronounced with a dull blue tint to them. Overall the linear version looks substantially better to me — the first image kind of screams “digital layering” (and that’s just so ’90s, you know).

As a little detour, this is a good opportunity to show what the Field mode in Conduit’s plot window is good for. Unlike Curve and Cube plots, the Field plot is only useful for the specific purpose of examining a transition effect. Here’s the Field plots for the video and linear dissolves:

The difference between the two maybe looks more enormous than it actually is. You have to remember that this plot is HDR; it is drawn in the range of -0.5 to 1.5 (as shown by the input boxes in the corners). Only the area fenced by the grey lines in the middle shows us the standard color range (0 to 1), the sides and corners are actually HDR blending. So the bottom-left corner shows the result of blending two negative input colors. The result is something negative, which is of course displayed in black. Likewise, the top-right corner shows the result when both colors are above 1, which displays as white.

Because we’re not dealing with HDR images here, we can just focus on the standard range in the middle of the field. In video mode, it looks just like we’d expect: a linear gradient from black to white. The linear version is more like a radial gradient with a rounded shape. The most important difference is seen if you look in the top-left and bottom-right corners of the standard range. At these points, the other image is full white (1) and the other image is full black (0). In linear light colorspace, this results in a much brighter grey than in video colorspace.

Well, enough with the Field plot… But please comment if you’d like to have a more detailed explanation of its workings, I can make a separate post about it.

The linear light blend is often going to have even more of an advantage if we do any color correction on the input images. On the right are the two blending paths used for the previous example, but with an extra Exposure node tossed into the mix on both sides. The Exposure node brightens the hands image so that its highlights are close to 1, and adds a bit of yellow tint.

The result images are below. This time the transition is at opacity 0.65. Rollover with mouse to see the linear version:

Click here to download the .conduit file for this effect

I know it’s not easy to get an idea of how the entire transition looks from just a few stills. I’d like to show a video comparison with some moving footage, but I’m running out of time now… So perhaps that will come later.