GI calculations from multiple cameras

[braveRabbit]

Is it somehow possible to gather GI information from multiple cameras per frame to reduce GI flickering in animations?

As far as I understand it only the currently rendering camera is used for GI calculations at this point. But, as we all know this can lead to GI flicker in animations (I am not talking about simple fly-through animations but moving objects such as cars and characters).

The idea would be to have multiple cameras in the scene, close to the camera which is used for the final rendering, but slightly offset. Different camera positions would cause different GI-noise patterns but they would all together be combined into one GI file and used for rendering the frame. IR & LC already produce cache files so it would be just a matter of processing multiple cameras and writing the information to one file.

The entire process would be done during the regular pre-pass phase of the current frame.

Is this possible? Or even a bad idea?
I just feel that in order to reduce GI flicker there is always a good work of tweaking involved and I find caching pretty annoying most of the time. Also the usage and workflow of the sphereFade node is not the best, because it needs multiple renders.

Any other ideas are most welcome.

Thanks.

[vlado]

It can be done, I suppose - you can render from all cameras with "incremental add to current map" mode or you can combine maps with the irradiance map viewer, but it will be tricky. But why not just use the "Use camera path" option for the irradiance map?

Best regards,
Vlado

[braveRabbit]

Thanks for your answer Vlado, still the solution seems a bit fuzzy.
The general question would be if "incremental add to current map" or "Use camera path" would even out the calculated GI result, meaning do these modes only fill in the gaps of the calculated map from previous cameras/frames (parts of the scene that get revealed in later frames and haven't been calculated before) or do they actually do some sort of interpolation to even out the GI result (this would be I am looking for)?

"Use camera path" wouldn't be working with animated characters, would it? Because it only would be written to one GI map and this would be counter productive with moving objects.

[svetlozar.draganov]

Both approaches will produce even GI solutions:

Incremental Add To Current Map will use the existing GI solution from the previous views and it will shoot new GI samples only in the gaps, Use Camera Path option will shoot GI samples from the entire path of the camera , so all the samples will be even out in the first place.

Use Camera Path could be used with animated objects as well but every-frame should have its own Irradiance-Map, it is not possible to use a single Irradiance Map for animation with moving objects.

[braveRabbit]

Thanks for the heads up, Svetlozar. Very informative.

Cheers, Ingo