Support multiple cameras

[cpetzold]

This is a first stab at multiple camera support. It currently assumes that the desired camera is on RenderLayer 0.

I'd like to allow flexibility on RenderLayer, but I wasn't sure how to support it properly. For example, if you have multiple render layers configured for an entity you want to render with bevy_vello, does that mean that you'd need to compute the affine for each camera targeted?

[simbleau]

Hey @cpetzold ! Thank you for the PR. One of us will review this soon.

cc: @seabassjh

[simbleau]

Ahh, good spot! We definitely copy/pasted this wrong ;)

[seabassjh]

if you have multiple render layers configured for an entity you want to render with bevy_vello, does that mean that you'd need to compute the affine for each camera targeted?

Yes. But if you're expecting that there will be n Vello scenes for n cameras, that will not be possible yet and requires something thinking and reworking on my render_scene system.

[seabassjh]

if you have multiple render layers configured for an entity you want to render with bevy_vello, does that mean that you'd need to compute the affine for each camera targeted?

Yes. But if you're expecting that there will be n Vello scenes for n cameras, that will not be possible yet and requires something thinking and reworking on my render system.

To clarify, this would have to be called n times
https://github.com/loopystudios/bevy_vello/blob/d25d787f14c77533b9b038ad153847b0f13e1c46/src/render/systems.rs#L181-L195

[cpetzold]

if you have multiple render layers configured for an entity you want to render with bevy_vello, does that mean that you'd need to compute the affine for each camera targeted?

Yes. But if you're expecting that there will be n Vello scenes for n cameras, that will not be possible yet and requires something thinking and reworking on my render_scene system.

Right that makes sense. I posted this on the bevy discord:

"I wonder if some sort of API where you add a VelloRenderer to a camera entity would make sense. A setup system would spawn the render target for that camera. Then it would create extracted instances per vello-rendered entity per overlapping camera with VelloRenderer."

Thoughts on that approach?

[seabassjh]

I agree on that approach. That was my initial idea as well!

[seabassjh]

What do you hope to accomplish with multiple cameras? Like split screen... because that would require us to revise the full screen triangle texture approach

[seabassjh]

I'm having a hard time thinking about what the practical uses of multiple "VelloRenderer" cameras would be. And what it would like look in a normal scenario-- multiple vello scenes overlayed right on top of each other on the full screen texture triangle?

[cpetzold]

I don't have a use case for multiple VelloRender'ing cameras. My use case is that I want to render an infinite 2d grid alongside my vello scene, and my approach is to use a separate camera to render a screen space shader.

[simbleau]

I don't have a use case for multiple VelloRender'ing cameras. My use case is that I want to render an infinite 2d grid alongside my vello scene, and my approach is to use a separate camera to render a screen space shader.

Ah, this is already possible with existing tooling.

In this, we're using egui + bevy_vello + a custom shader (infinite "transparency" grid).

This was all Sebastian's doing, so I'll wait for him to talk about how he did that.

[cpetzold]

I did get it to work with a single camera approach, but it felt a bit bad because I had to scale the shader mesh with the camera to support zooming the camera. Though it does simplify the shader code since I can just use the View that's automatically passed!

[simbleau]

@seabassjh

[simbleau]

Merge conflict @cpetzold

[simbleau]

Superseded by #68