Question Cinemachine collider doesn't work like typical third-person game cameras

Unless I'm missing something, it seems like it's not possible to use Cinemachine Collider to implement a standard third-person game camera. This is a bit difficult to describe and perhaps a little subtle, so hopefully this explanation makes sense.

Typically, a third-person game camera will prevent the camera from clipping into objects by moving itself forward – which is possible with Cinemachine Collider. However, most third-person games do this while still allowing you to place an object in between the character and the camera.

For example, imagine a character standing near a tree. In games like Zelda: Breath of the Wild, orbiting the camera around will allow the tree to pass between the camera and the player without the camera zooming in. However, if you back the camera into the tree, the camera will get closer to the player, still preventing clipping.

Playing with Cinemachine Collider I can't get anything that feels even close to right, because anything blocking the raycast to the player causes an awkward zoom in.

Perhaps there's a way to accomplish this out of the box – I'm not sure what to call it so searching hasn't been fruitful. Does anyone know if that's possible, or if there are any resources that describe an optimal third-person camera setup in more detail?

 

A painful alternative hack may be to disable colliders of terrain around you so the camera can move through it as the player gets close to it. Or even implement a raycast from the player in the direction of your camera and items it finds that are terrain it disables the collider for and then when you move out of the area, it enables them again.

 

Thanks for responding!

I'm pretty sure that this behavior is very similar in every third-person game I can think of (Hitman, Uncharted, Assassin's Creed, etc.), I can try to gather more examples. My goal is simply to have a well-behaved camera like those games, I'm not 100% sure of the implementation specifics. Because it's so common I think it would be awesome if it was built-in to Cinemachine, but I'm happy to hack something together.

I believe that this behavior applies to all objects in the game, based on some testing.

Here's a recording of the camera in Zelda

Here's my third-person camera WITHOUT cinemachine collider

Here's my the camera WITH cinemachine collider

You can see that without the collider, obviously the camera will happily clip into whatever, as expected.

However, with the collider, getting an object between you and the camera results in unexpected, jarring, and ugly behavior when line-of-sight is broken because it zooms in way too close. I can't think of any games with this style of third-person control that do that, so it seems like the current collider implementation is optimized more for cinematics or other control schemes.

 

My guess at the implementation is that the camera has a collider on it but NOT a raycast, and if the camera bonks into something it will get closer to the player. If there is an object between the player and the camera you can move away from the camera and "pull" the camera through the object, but if you back up the camera will hit the object and cannot be "pushed" back through it. (This is specifically demonstrated in the Zelda example video.)

 

Oooooo good call! I've set my "Transparent Layer" to "Everything" and it seems to mostly work like this! I am still seeing some occasional clipping but maybe some of the other settings need to be tuned.

 

Setting it up with everything on the Transparent layer definitely works better than before, but there's still some definite weirdness. In particular, there are lots of discontinuities when orbiting, but I'm not sure what's causing the camera to move like this. Here's a video showing the discontinuities (I'm playing with a controller and orbiting smoothly in this clip.)

 

Here are the collider settings I'm using:

 

Hi wilgieseler,

What is your basic camera setup? Can you post a screenshot of that?

Cheers!

 

Trying now to just exclude smaller objects via layers/tags, but still can't figure out why the camera is translating weirdly when occluded instead of pulling forward. Here's a screenshot of most of the camera settings.

 

Been playing around improving the settings and I think "Camera radius" is what causes the discontinuities.

 

Oh those collision debug lines are very helpful! Still tweaking things, trying to reformulate my question since I'm trying the recommended approach of simply manually specifying which items the camera can collide with. Not how Zelda works, but seems that Hitman works similarly to that with pretty good results.