Author Topic: Progressive render and change detection in TeamRender  (Read 2246 times)

2018-03-27, 10:20:05

Rhodesy

  • Active Users
  • **
  • Posts: 553
    • View Profile
Im just listening to the latest CGGarage podcast about Vray cloud https://www.chaosgroup.com/cg-garage and it sounds amazing. But it makes me realise just how far away from this sort of thing we are on a local level with TR!

They are describing how after initial upload it just detects changes in the scene - like IR does and that you can also use progressive rendering which means you can check your scene after a few passes to see if everything is OK. Basically its like final frame rendering with IR features. I imagine its the same for rendering flythrough frames where 99% of the scene doesnt change. I dont know if on a local level, Vray Swarm is quite there yet but I'm sure this tech will roll out across all vray platforms soon. And they support all the common plugins.

I dont know what is coming with Team Render but I cant help but think it is a limited system and reliant on Maxon.

Our current option is the 'scanline' style packets of build up where a pre-set number of passes is reached and there is no denoising available. This means we might not be able to spot any mistakes until 3/4s of the way up the image and we have to render for longer as we dont have denoise (but I know this feature will come). It can also take ages to package up and send out over the network and if you are wanting an animation it has to regenerate the whole scene each frame which can easily take 5-8 mins before it starts rendering so its really prohibitive to use. We've just bought a 10GBe network (yet to set up) so I'm hoping that is going to help but it will still require the whole scene to be set out and fired up each time.

I have never used the max version of corona - is progressive rendering available there on distributed?

I know there is a strong argument for Team Render and compatibility but I really wish there was an equivalent of Vray standalone / swarm and I would happily work within a set number of provided corona shaders that covered the major features like randomisation and a handful of key noise shaders. Ultimately I would love a universal material format so we could use materials created in .max for bought assets.

Sadly it all feels like a long way off. Even when we get the feature complete TR version with denoising and noise thresh its still not an ideal solution for using on a daily basis mainly due to scanline style preview and time taken on making changes / restarting the render which can take an age on a big scene. Is there any tech you can collaborate on with ChaosGroup now you are all one?

Thanks