Basically, I want to try and render an animation faster by maybe decreasing the samples or light bounces.
Right now I have a scene that would take 15min to render per frame. But I want to have a character do an animation in said scene for 500 frames. So I’m interested on decreasing these render times.
So I know that some people bake their scenes to render them faster, so would this work correctly for an animation with a character? And if it is, what is the correct way of using a baked scene?
I just tried using my baked textures on an emission shader, but I still need those high samples. Which defeats the purpose of baking the scene I think.
I would appreciate to have some information about this from people that have used this or use this often.
You need to identify what causes the high sample requirement. What does the scene look like, what’s the lighting involved, are there any tricky material setups (aniso shading can be notoriously noisy)? Are you using denoise? Can some of the noise (in render passes) be handled in post? Have you done anything to prevent noise (i.e. clamping)?
What do you mean “baked textures on an emission shader”? You need to bake the lighting information. I.e. big and small light sources combined can be horrible together but fine in isolation. So you can either render out in isolation and add later, or render out the problematic light (usually the small ones) and add them back as emission shader while still keeping normal response from other lights.
Not trivial to setup, and it’s not something I have ever done (except render out different light passes). I don’t really bake out general lighting myself.
renderstreet (online render farm) is your closest solution. You´re gonna have to deal with speed times at your own, once you get good with light and materials and animation.
Blender cannot (at the moment) be speedier than it is using cycles. This is why EEVEE was introduced. Even so, it will still take a good graphics card to render to (at least) hd quality.
Consider online render farms.
I’m going to stick to EEVEE or UE4 then. I don’t feel confortable paying that ammount of money for a little side project that has no financial gain to me.
There are various things that you can do. For instance, if the camera isn’t moving, you need just one frame of the background. You can individually render each character, and perhaps also the shadows cast by that character. You assemble the final animation in the compositor.
In addition to saving time, this approach also gives you more control. You can “tweak” the various elements as you combine them together, and this is a very fast process because it is now two-dimensional. (If you use MultiLayer OpenEXR files, you can capture multiple “layers” of information such as Z-Depth and use this in the composite step.)