Loading...
> >

Any tips for rendering large scenes? Is there a way to preserve lighting while hiding an environment

ParadigmParadigm Posts: 423
in The Commons

Hi all,

I'm about to begin one of my largest attempted scenes with 7-8 fully clothed Genesis 8 characters in an indoor scene and I don't think even a 2080TI will hold all of that (but I will certainly try!) so I'm looking for some tips from those who have experience rendering scenes larger than the card's VRAM allows.

My first thought is to render the background first and then hide it for the character renders. I've attempted this before, but the issue is when I hide the building or whatever they're in the HDRI casts much more light. Is there a way to hide the scene from rendering but still have it cast shadows?

Or should I just hide groups of characters at a time and merge the layered renders?

Thanks!

Comments

  • scorpioscorpio Posts: 8,533

    If you have this

    https://www.daz3d.com/scene-optimizer

    it can really help.

    Depending on the scenery and the clothing and hair you use 7-8 on a 2080ti may be possible, I've got 8 fully clothed and scenery on a 1080ti. 

  • ParadigmParadigm Posts: 423
    scorpio said:

    If you have this

    https://www.daz3d.com/scene-optimizer

    it can really help.

    Depending on the scenery and the clothing and hair you use 7-8 on a 2080ti may be possible, I've got 8 fully clothed and scenery on a 1080ti. 

    I do have it but I don't like using it. I travel for work and sync my data to a harddrive so I can also render on the road and the optiomizer stores the files in a location that I don't sync so it messes with the scenes. I can get it to work but its enough of a pain in the butt that I prefer to simply batch render than deal with it. It was amazing when I had a smaller card and only rendered on the one computer though. Definitely a great product.

  • MelissaGTMelissaGT Posts: 2,611

    Sounds like using canvases and then compositing them with postwork is your best bet. 

  • AsariAsari Posts: 703
    edited April 2020
    If you have a lot of characters then probably not all of them are close to the camera or in focus. You can turn those character resolution to base and reduce texture size, remove normal maps that you don't need, and so on.

    Or you could render twice, this works well with DOF on: render the blurry background in relatively low resolution. Render the objects and figures in focus in original resolution with spit rendering and delete all figures and objects from the spot rendering area. Then put both parts together.

    Post edited by Asari on
  • ParadigmParadigm Posts: 423
    melissastjames said:

    Sounds like using canvases and then compositing them with postwork is your best bet. 

    I've never used canvases, but looking at a quick tutorial I think that's exactly what I need. I can render the environment with the characters on a different canvas so they'll cast shadows but not be in the scene and then do the reverse. The real question is if the items are still loaded into memory when using canvases. I imagine the geometry is to a degree in order to cast shadows but not the textures at any significant level of detail.

  • AsariAsari Posts: 703
    Sometimes canvasses yield different results ... I know in my renders when I use strong rim lighting the hair in the beauty canvas looks different when composed together with the background. Your scene is different, try if it works in your scene.
  • SevrinSevrin Posts: 6,313
    Paradigm said:
    melissastjames said:

    Sounds like using canvases and then compositing them with postwork is your best bet. 

    I've never used canvases, but looking at a quick tutorial I think that's exactly what I need. I can render the environment with the characters on a different canvas so they'll cast shadows but not be in the scene and then do the reverse. The real question is if the items are still loaded into memory when using canvases. I imagine the geometry is to a degree in order to cast shadows but not the textures at any significant level of detail.

    Canvasses are for separating components, primarily lighting souces and effects, in a single scene.   You can use them to split parts of the scene onto different canvasses, but won't solve the problem you're having here.  They won't allow you to have anything more visible than rendering without.  For what you are trying to do with fully textured assets, you will need to do multiple renders, hiding some parts of your scene in one pass and other parts in one or more other passes.  Whether you use transparent PNG files or Beauty canvasses with alpha enabled is your choice.   You can always add shadows by hand in your image editor whichever way you choose to go about it.  In fact, you will probably need to.

    Canvasses are great, but I would recommend learning about their use with a simpler scene.

Sign In or Register to comment.