Blackmagic Forum

I'm trying to do a simple screen replacement in Fusion and am having a lot of trouble. Details:

• The source shot is a zoom out of an old television screen
• We did not have any sort of green screen or tracking points on the television screen. Instead, we figured we could just mask out the screen itself (there are clearly defined edges where the glass meets the edges of the TV frame) and do a screen replacement
• Because we did a zoom, there is a certain amount of tracking for the mask that is required
• We also shot at a roughly 30 degree angle from the front of the TV screen, so we need to add some skew/distortion to the image on the television screen

Our goal is for somethign that actually does look a little fake/cheesy, as that's the vibe of the video. However, we obviously want the image on the TV to look like it belongs there.
In my mind, this is relatively easy:

1. Create an mask around the edge of the TV glass (it's an old CRT tv, so there are rounded edges
2. Track this mask to account for the zoom
3. Place the "TV footage" underneath so it shows through the mask
4. Transform/warp the TV footage so it matches the perspective
5. Attach TV footage to the track

However, I'm still a relative newbie to Fusion and am having a hard time with this. Most YT tutorials I have found related to planar tracking are for placing things on TOP of footage (like text), rather than masking to reveal something underneath. Screen replacement tutorials are all geared towards using green screens and tracking points or don't required rounded edges, so again, their screen footage is foreground and their original footage is background.

It seems like mine needs to be:

1. TV as foreground with mask for TV screen/glass
2. TV footage as background to appear through the mask

Really feel like this should be simple, so I'm clearly missing something.

For screen replacements I usually plus the footage on top of screen, so that all the original reflections etc are kept also. Screen surface is not uniform, it reflects the environment, and usually you want to keep that interaction. As screen emits light, plus operator is the one to use. I'd mask the tv footage and then do the merge, don't see any benefit in placing it under, and mathematically it produces the same result. But if you want to do it a bit cheezy, anything goes

Hendrik Proosa wrote:For screen replacements I usually plus the footage on top of screen, so that all the original reflections etc are kept also. Screen surface is not uniform, it reflects the environment, and usually you want to keep that interaction. As screen emits light, plus operator is the one to use. I'd mask the tv footage and then do the merge, don't see any benefit in placing it under, and mathematically it produces the same result. But if you want to do it a bit cheezy, anything goes

Thank you, I will give this a shot. Yeah, I was lying in bed last night thinking that I should try to keep some of the screen detail/reflection itself.

You mention the Plus operator, but I only see Over, In, Held Out, Atop, and Xor in the Merge node?

Check out this tutorial: (the Fusion PDF for Fusion 15)
https://www.blackmagicdesign.com/produc ... e/training

Work through the van logo planar tracking exercise. If you're new to Fusion I strongly recommend you start from the beginning.

I also started on YouTube and paid the price with learning wrong techniques and having to fix that.

Travis Ward wrote:You mention the Plus operator, but I only see Over, In, Held Out, Atop, and Xor in the Merge node?

There are two ways to do an Add:
1) Use the ChannelBooleans node instead of Merge, but don't forget to set Alpha to Do Nothing, or you wind up with an alpha of 2.0, which can have unfortunate results.

2) In the Merge, set the Alpha Gain slider to 0. That eliminates the effect of the Alpha on the Background input, resulting in an Add operation.


If you want the gory details of that second one, read on.

An Over operation, in its simplest form is this equation:
FG + BG * (1-A)

where A is the Alpha channel of the Foreground input. This assumes that the Foreground is premultiplied by its own Alpha. If it isn't, the Additive-Subtractive slider in the Merge changes the formula to this:
FG * A + BG * (1-A)

The Alpha Gain slider (g) is a multiplier on A, resulting in this:
FG + BG * (1-A*g)

So if g = 0, the formula simplifies to:
FG + BG * (1-0) = FG + BG * 1 = FG + BG

If I recall correctly, the Merge formula is further detailed in the Fusion Tool Reference manual.