Z-Depth and Zooming

[Hoptoad]

I've been working on a tracking shot. It starts with the camera aimed at the ground, focused on a bug. The camera tilts upward as the bug climbs a branch. Then the camera tracks the bug left to right as the bug crawls along the branch. The distant landscape has z-depth for the layers, for parallax scrolling.

... Don't move the camera!!!

You might think that what is the harm if I move the camera? A lot. Everything moves when you move the camera. That almost prefect animation that your halfway through will get ruin by moving the camera. Trust me. I spent tons of time fixing this mistake.

This quote makes sense to me. Every adjustment of the camera changes everything in my shot due to z-depth. Sometimes the bug would be in the center of the frame like I wanted, but after adjusted the zoom or tilt or pan, the bug might not be in the frame at all.

This is my first serious attempt at using z-depth. The layers are arranged thusly, from closest to farthest:
- bug
- branch
- ground
- trees
- hills
- mountains
- sky

Any tips on using z-depth? What numbers do you tend to use for generic landscapes? Do you have a system, or do you just wing it? I activated the orbit camera, then put the cursor in the z-depth box of the each landscape layer and used my mouse's scroll wheel to change the number by a small amount. I think I'm not moving the layers far enough apart...

If I were to start over and simply move the layers instead of moving the camera, any tips for creating the cool parallax scrolling effect? I know further layers move slower, closer layers move faster, but anything else?

Another question: How do you do a dolly zoom in Moho? I kinda want to do a dolly zoom on the bug.

Another question: What is you opinion on this shot: super zoom (I'm not sure what to call it). In the center of the frame is a house, on a far away hill. The house's window is tiny but nonetheless visible, and attracts our attention by its color. Suddenly, the camera zooms-in on the window in the space of one second, then you (the viewer) are through the window and inside the room. Is this shot disorienting, in a bad way?

I created a super-zoom shot in my cartoon and I'm not sure it works. On one hand, it looks pretty cool. SUPER ZOOM! But on the other hand, it's almost like you (the viewer) flew like a rocket in one second toward the window, and now you are through the window and inside the room.

Would you personally use a shot like that -- or have you? Or would you simply cut to a close-up of the house, hold it for a second, then cut to the inside of the house?

It's such an important moment in my cartoon. I need the viewer to know that the next scene occurs inside the house in the distance. But maybe I'm not giving the viewer enough credit; maybe the super-zoom isn't necessary. Maybe it's obvious that the next scene occurs inside the centrally located house on the hill.

What would you do to connect the two scenes? Use a super zoom, or something else?

Thanks.

[synthsin75]

Dolly zoom is just like in real life. The camera is tracked closer or farther while the camera zoom is opposite. As long as you have foreground and background separated in z-depth, it should work.

For your "super zoom" shot, I'd be apt to do a slow zoom, only a handful of frames, that fades to the interior, unless there's some reason it needs to be quick or dramatic.

[funksmaname]

Hey Hoptoad,
Not sure if you're aware as you didn't mention it.
Using the 'layer transform' tool - if you hold alt and shift while moving a layer, you will notice it get smaller or bigger but then snap back to its original size. If you orbit the camera you will see what this is doing is moving the Z axis while maintaining the visual scale so when you plan a shot, it will look identical before you start moving the camera...

Now, as far as super zoom... I'm not sure its necessary, even if you have a small house in the distance and you do a very slow move towards it then fade into the house or something I think the audience will get it as long as the design of the landscape isn't too cluttered. But if you want to zoom over a long distance, i would probably ease in and out quite a lot, so only the middle of the zoom is fast... and use motion blur. I dunno, it depends on the style of your film and what you're trying to achieve, and how long you want to spend establishing the shot and zooming.

Not exactly the same but I recently made a video with a similar establishing shot. I think it would still work if the house was smaller, and you went straight to the interior. The entire video had to be like 12 seconds so didn't have long outside... I think lead the eye with contrast and you don't need to take the viewer on a rollercoaster

[synthsin75]

I agree with Danny, at least on easing into any fast zoom.

[Hoptoad]

I found a video on YouTube that called such a shot a crash zoom, so I should call it that instead of super zoom.

Okay, I'm going to change the crash zoom in my shot to something else, probably a transition like the one in the post above. The crash zoom was too visually jarring. Tarantino uses crash zooms a lot, but for action scenes, usually. Rapidly zooming into a distant house where a guy is sitting in a chair is not exactly an action scene.

Thanks for the feedback.

[hayasidist]

Any tips on using z-depth? What numbers do you tend to use for generic landscapes? Do you have a system, or do you just wing it? I activated the orbit camera, then put the cursor in the z-depth box of the each landscape layer and used my mouse's scroll wheel to change the number by a small amount. I think I'm not moving the layers far enough apart...

I could spend an hour or more discussing how to map moho coordinates to "the real world" - but I won't.

Bottom line for me: you can (try to) use z to simulate real world distances between (e.g.) the camera, your bug and the mountains - so with the camera default z at 3.732 with a lens equivalent to a 45mm focal length in a 35mm fullframe stills camera [there's a rational explanation for those numbers in the "long version" of this discussion] and the height of the moho render window at exactly 2 units you can run the maths and get a number that says 1 moho unit = 1 metre (or whatever). which kinda works right up to and excluding the point that you realise that you draw "the ground" on a vertical (constant z) x/y plane - and because you know about "these things" you draw it with proper perspective so (e.g.) the top of the frame looks to be "10Km away" and the bottom of the frame looks to be at somewhere near the feet of the person holding the camera -- in other words the vector layer at z = 0 actually covers terrain that stretches far into the real world (z) distance. You try to get round this by rotating the layer around the x axis then realise that you need the drawing height to be 10000 moho units if you want it to be 10Km away... and the bug needs to be drawn at 0.01 moho units (1cm).

This from 8 years ago: https://www.youtube.com/watch?v=PjFulEo2auw used that approach - together with camera zooms / tracks / ... It was primarily done as an experiment to see how hard it would be to create and move around a 3d environment in a 2d program... I haven't ever used that approach again.


Having found the proverbial brick wall and bashed your head against it you decide to do what the matte painters in "real movies" did and have layers that seamlessly join together from where the camera will be looking at them (hence the "don't move the camera" advice). And z becomes little more than a "stacking order" for use in a "traditional multiplane camera" ... unless you want to get into using moho's depth of field... but that's another story.

[slowtiger]

Just some numbers I used for this "3D" setting:


The buildings have these Z values:
0
-0,5 (left and right side of the street alternatively)
-1
-1,5
...
-10
-11
-12
...
-25

Sky: -50

Camera uses a zoom value of 20, Z starts at 5, ends at -17, and uses rotate and tilt during the flight.

[Hoptoad]

Thanks, slowtiger, I was hoping to see some numbers posted. I suspected I was being too restrained with the z-depths, as I had only moved layers a fraction of a number in most cases.

This from 8 years ago: https://www.youtube.com/watch?v=PjFulEo2auw used that approach - together with camera zooms / tracks / ... It was primarily done as an experiment to see how hard it would be to create and move around a 3d environment in a 2d program... I haven't ever used that approach again.

Having found the proverbial brick wall and bashed your head against it you decide to do what the matte painters in "real movies" did and have layers that seamlessly join together from where the camera will be looking at them (hence the "don't move the camera" advice). And z becomes little more than a "stacking order" for use in a "traditional multiplane camera" ... unless you want to get into using moho's depth of field... but that's another story.

Z-depth is tricky. In my limited experience with z-depth, I discovered that animation became much harder if I tried to combine two camera movements. I can simply pan or track or zoom or tilt and get good results, but when I combined two or more of these movements, it was hard for me to predict the outcome and hard to correct mistakes.

Thanks for mentioning the good ol' multiplane camera! I decided that I will embrace the aesthetic of the traditional multiplane camera for my next complicated tracking/zoom shot -- in other words, move the layers and leave the camera stationary.

unless you want to get into using moho's depth of field... but that's another story.

Whoa, I just watch a tutorial on depth of field. Thanks for mentioning it. Depth of field looks awesome. It seems like depth of field is what I actually want for the initial shot with the bug, not dolly zoom. You (the viewer) are looking at the ground. The bug is between you and the ground, on a branch, and is blurred and not exactly noticeable. But then it moves, attracting your attention, so then the camera focuses on the bug and the background gets blurred. That really sounds like depth of field.

[Stan]

This thread reminded me that I have an old tool that does pretty much the same as Alt+Shift+Drag of the LM_TransformLayer tool, but also allows you to enter the Z-value by typing in the number. You might find it useful when setting up your scene in 3D space. I just published the tool here: https://mohoscripts.com/script/sz_z_tra ... auto_scale

[slowtiger]

I should add that it helps to watch your setup in orbit view. As you see in my example, the buildings are pretty much placed as in a real street. The sky is that large to prevent any border showing when tilting the camera. Keep in mind that 1 = one screen height. Usually I build setups like this just as "looks good enough", but you may encounter a scene which requires more planning. In that case create a bitmap map or ground floor plan, import it, tilt it 90°, and adjust your planes in orbit view to that.

I remember that I did a basic camera movement here, then placed more buildings. If I want a certain view at a certain moment of the flight, I adjusted elements there (enter Z value first, scaling and positioning later), then copied those keys to frame 0 and erased the later ones.

Another tip with these setups: first I create a group layer with more group layers inside, those get named by Z value and get that Z value. This way I can easily put images inside those subgroups and have them placed in Z depth automatically.

Last tip: I create a group layer setup with Z values from -1 to -10. Then I duplicate the whole group and set its Z to -10, which puts the layers inside at -11 to -20.

[Hoptoad]

This thread reminded me that I have an old tool that does pretty much the same as Alt+Shift+Drag of the LM_TransformLayer tool, but also allows you to enter the Z-value by typing in the number. You might find it useful when setting up your scene in 3D space. I just published the tool here: https://mohoscripts.com/script/sz_z_tra ... auto_scale

Thanks for the script, Stan. I tried it on my "bug layer", but the results surprised me. I expected the bug to remain where it was while the z-depth changed, but the bug immediately left the workspace window, probably because of the other camera movements already in place. I'm guessing the script should be used prior to animating the tracking camera, not after. I imagine the script would be very helpful during the early stages of setting up a scene, when the z-depth of layers is being decided.

I should add that it helps to watch your setup in orbit view. As you see in my example, the buildings are pretty much placed as in a real street. The sky is that large to prevent any border showing when tilting the camera. Keep in mind that 1 = one screen height. Usually I build setups like this just as "looks good enough", but you may encounter a scene which requires more planning. In that case create a bitmap map or ground floor plan, import it, tilt it 90°, and adjust your planes in orbit view to that.

That's a great tip.

Seeing your buildings placed like on a real street via orbit-view made me realize when I altered the z-depth of a layer by a minuscule .2, the amount was woefully inadequate and a huge chunk of the reason I was having trouble getting the parallax effect to work like I wanted.

I didn't realize "1" equals one screen height. Thanks for mentioning it. I was curious why it wasn't zero.

I remember that I did a basic camera movement here, then placed more buildings. If I want a certain view at a certain moment of the flight, I adjusted elements there (enter Z value first, scaling and positioning later), then copied those keys to frame 0 and erased the later ones.

Another tip with these setups: first I create a group layer with more group layers inside, those get named by Z value and get that Z value. This way I can easily put images inside those subgroups and have them placed in Z depth automatically.

Last tip: I create a group layer setup with Z values from -1 to -10. Then I duplicate the whole group and set its Z to -10, which puts the layers inside at -11 to -20.

I'm going to do those things. They seem time-saving. The more I animate, the more I realize that I need to get everything right as fast as possible.

Thanks.

[striker2311]

Just some numbers I used for this "3D" setting:


The buildings have these Z values:
0
-0,5 (left and right side of the street alternatively)
-1
-1,5
...
-10
-11
-12
...
-25

Sky: -50

Camera uses a zoom value of 20, Z starts at 5, ends at -17, and uses rotate and tilt during the flight.

Emmm!!! Can i ask you how did you put that flash holding effect in the last It's quite good

[slowtiger]

I don't get what you mean - could you rephrase it? Or give a time signature?

[striker2311]

I don't get what you mean - could you rephrase it? Or give a time signature?

Sorry it was a typo I mean flush not flash

Time signature -0:25

[slowtiger]

Ah! You mean the bubbling water? That's just an old-fashioned practical effect.

I have a rectangular glass vase, about 8 x 12 x 20 cm. Put it on a table in front of a black BG, set a strong light from one side, my little cheap video camera in front of it. Poured water from about 30 cm height into it, filmed that. Reversed the footage, increased contrast. That's it. (Maybe I sped it up a bit as well, don't remember.)