How to resize only one object in a video?

Hello,

I have a customer who wants me to shoot their music video and in someone else's video they saw a special effect where certain items got larger in the video without anything else getting larger as well. For example in one scene the artist had a chain around his neck and it grew to about 3x the original size then shrunk back to normal size.

I have tried using the BSpline tool combined with a planar tracker and a transform node to try to mask out a single object then I used the transform node to resize the masked out object and I added the same video track as a background in the merge node to reinsert the originally sized footage but it turns into a mess.

Depending on the angle and how much camera motion and subject motion there was, the original object may start to show behind the larger one in front.

I know I am missing something simple...it seems like I just need to mask out the object and resize it but the mask and the resized object don't interact naturally with the rest of the video when I do that. But when I add back the original video track then the two don't interact well either.

It is so hard to find good Fusion training material, I don't know how anyone gets truly proficient at Fusion without hours upon hours of practice.

You need a bunch of good quality roto to isolate the object you are interested in and also get occlusion mattes for objects that interact with it or cover it. For example if you have chain around the neck and you want to resize it 3x, it might be big enough so that when artist moves his chin it must cover the chain when in original footage chain was small enough to not be covered etc. So loads of roto and comping is ahead for you. If you want to do do it nicely you must create new shadows too because bigger objects cast more shadow. And to get the movement, lots of fiddling with stabilization and tracking. Because you want to use the original motion but object itself is scaled 3x so original motion is also scaled. Stabilize the object, rescale, reapply motion and fixes.

Thank you Hendrik I guess I am on the right track then. The effect looks so simple I thought I was missing something simple. I will read up on rotoscoping and occlusion mattes. I will also try to talk the artist out of using the effect; I doubt they have the budget necessary to pay me to spend hours on creating it.

I don't know how anyone gets truly proficient at Fusion without hours upon hours of practice.

I don't know how anyone gets proficient in *anything* without hours upon hours of practice. Isn't that what proficient means?

Back on topic: can you show the video where your client got his inspiration?
There's a good chance that this video was planned in advanced so that the effect was actually doable.
First filming something only to decide later you want to do something like this is usually an expensive recipe for disaster.

I don't know how anyone gets truly proficient at Fusion without hours upon hours of practice.

I don't know how anyone gets proficient in *anything* without hours upon hours of practice. Isn't that what proficient means?

Back on topic: can you show the video where your client got his inspiration?
There's a good chance that this video was planned in advanced so that the effect was actually doable.
First filming something only to decide later you want to do something like this is usually an expensive recipe for disaster.

Yes, that was poorly worded, what I meant is that there seems to be so few good training resources for Fusion. Most of the YouTube videos are about basic effects like working with titles and basic Davinci Resolve editing, I haven't found much really good training on Fusion so far. AfterEffects by comparison has example videos of literally anything you can imagine so you don't have to spend hours experimenting, you have a good starting point on what you want to do pretty quickly.

Fusion is incredibly powerful but I am not a motion graphics artist, I simply do not have the time to experiment with every node; my way of quickly getting a finished product is usually to find an instructional video that gets me around 60% of the way there and by then I can usually figure out the rest.

Here is the client's reference video, I apologize in advance for the lyrics and content, I do not accept clients that want videos like this (mine doesn't) but he did like the enlarge effect that was used in the video, you can fast forward to 00:38s to see the first time the effect is used. The first few times look pretty easy, but then he enlarged his head, the shoes, etc. and that was during complex subject and camera movement which I would have a hard time duplicating:

Well. In this clip there wasn't that much planning in advance either. For some of the shots yes. (enlarging the tongue with CGI) but a couple of the enlarged heads are very clearly cut out of the original footage.
The mattelines around the hair are very obvious. Also most of the shots last maybe 25-50 frames tops, so it's less work to get it done. So manual labor is your only option I'm afraid.

Keeping shots short definitely helps because viewer has no time to grasp the errors, only get a glimpse of the general effect.

Hey,

The average of what I saw in this clip is not reaaaallly tough actually. As Hendrik or Sander, I supposed they may have had separated graphics or clips of what grew bigger, but now I saw it (heads, tongues, ...), I think you'll make it with a simple roto + track + warp.
So find the right selection you need, get it aside the way you prefer in a single pic frame (with a timestretcher node maybe), track the real (let's say head in our example) head and merge it back on the clip after a keyframed transform node, or a warper one just for you to see what it makes...
That's just the basics of the global animation you need, you can tweak (matte control, etc etc)

Thanks everyone and especially Sam for the details, some of which went over my head. I think I talked the client out of it and told him if he really wanted it I would outsource and he'd have to pay a motion graphics artist.