Camera Blending: Keep Distance while Blending

Hi,
I have a bunch of VCams positioned around a single target. They are toggled by user input. Now when blending from one cam to the next, position and rotation are interpolated linearly. This means when the current cam is on the opposite side of the target, the camera moves through or extreme close to the target.
Is it somehow possible to take Camera Distance (each vcam has different value) into account while blending from one cam to another? Or is there a better way to accomplish something like this?

 

Hi @zyzyx Interesting scenario. Yes the standard blends will 'go straight' to the next camera. I get you're trying to preserve an arc.

First thing, I can see you're on an older version of CM. Best to grab the latest HERE and if you're reading this a day or two after this was posted, the final released version should be on the Asset Store. Either way, get onto cm v2.1

An approach you might want to try is using an Orbital Transposer Body method on your vcam and then animating / code driving the Heading Bias. That will cause your camera to rotate around its Follow object - which you can set to be any object at a desired orbital point.



You can poke the Heading Bias with this: m_Heading.m_HeadingBias

Give that a try and let us know how it goes

 

Hey @Adam_Myhill
Thanks for your quick reply! I will try what you suggested.

Quick followup question: Are there any plans to let us plug in a custom vcam lerp function in a future update?

 

Not immediately on the horizon. It seems you want to do a form of orbit so try the Transposer Heading idea.

Or, why not create a dolly track with exactly the shape you want?

 

Well, there are many camera positions (even user created ones) so I would have to make sure there is a dolly track from every cam to every other cam.
Will try the Transposer Heading, thanks again!

 

Ah gotcha. I see now why you're interested in different blend strategies.
Try the Transposer Heading, I have a feeling it will sort of work but not between user generated cameras.
Let's discuss this further and do so after you've given that a try so we're on the same page in terms of options, or partial solutions. Cheers.

 

So I tested the Transposer Heading idea but it ended up being too restrictive and I would have to animate the target as well.

I ended up modifying the Lerp method in the CameraState.cs and changed:

Code (CSharp):

1. state.RawPosition = Vector3.Lerp(stateA.RawPosition, stateB.RawPosition, t);

to:

Code (CSharp):

1. state.RawPosition = Vector3.Slerp(stateA.RawPosition, stateB.RawPosition, t);

Pretty simple but does the job for the most part. It does get ugly in some edge cases though. I will try to tweak it a little bit to make it more robust.

 

Solid idea, we'll give that a look here to see how it responds.

There's definitely a good idea around the ability to select different lerp methods... On a waaay older predecessor to Cinemachine we had a number of options available. In theory it was great, in practice most of them were hardly used.

 

After working with the above solution for a bit I think a spherical/cylindrical interpolation option for blending VCams would be very useful for scenarios like a garage in a racing game where the player can tune and view the various parts of a car, or a spaceship hangar, etc.

 

I tried doing it in LateUpdate and it worked well, but ended up modifying the CameraState as described above (which is much more convenient for what I'm doing).

Code (CSharp):

1. //state.RawPosition = Vector3.Lerp(stateA.RawPosition, stateB.RawPosition, t);
2.  
3. // "cylindrical" lerp
4. var v_up = Vector3.up;
5. float state_posY = Mathf.Lerp(stateA.RawPosition.y, stateB.RawPosition.y, t);
6. var tempPos_A = Vector3.ProjectOnPlane(stateA.RawPosition, v_up);
7. var tempPos_B = Vector3.ProjectOnPlane(stateB.RawPosition, v_up);
8. state.RawPosition = Vector3.Slerp(tempPos_A, tempPos_B, t);
9. state.RawPosition += v_up * state_posY;

 

Hey thanks for the heads up! This sounds awesome!