Octane for Telivision Serial currently under production

[RajivMudgal]

We are a using a combination of Blender and Octane for all our rendering needs. They work well for episodic television series where Budgets are restricted and time-management is paramount. We experimented with various animation and rendering software searching for that elusive compromise between cost, time and quality and ended up choosing Blender and Octane as our obvious choice for this project.
These days we are sticking everything into Octane and are loving it.

We have also Posted some Animatics and early renders. They can be found here: https://www.youtube.com/channel/UCmhW4B ... 48w/videos

[suvakas]

Very cool !
What's your rendering hardware and how fast are your rendering times?
We are planning to build something similar on top of Octane.

Suv

[RajivMudgal]

Hi
For characters animation, we are using a single dedicated system for all our renders.
Its equipped with 4 Nvidia Titans.
It usually takes 2 to 3 render minutes per frame.
And the ZPass takes around 30Sec.
We render at 2K and sampling is set to 2000, we also do some post noise reduction when needed.

For backgrounds, environment and other things we use the internal renderer as we end up using a lot of fluid dynamics and particles to terafarm our worlds. We have 22 Dedicated Cpu's for that task.
It takes around 5 to 6 minutes to render per frame.

[suvakas]

I see. Thank you for the info.
Very reasonable rendering times for such resolution.

Suv

[JimStar]

RajivMudgal
Just in case: don't forget that you can dramatically increase the Octane animation rendering speed if you properly set all mesh types ("Global", "Scatter", "Movable...", "Reshapable...") and set animation mode to "Movable proxies". In this case only the "movable" objects will be reloaded every frame - only transforms will be reloaded for "Movable" types (it's very fast), and only for "Reshapable" types the full mesh will be reloaded and evaluated every frame. So, this type of tuning will speed-up the "translation" part of every frame's rendering.
For non-movable objects which are not scattered - the "Global" type will increase the rendering speed itself (samples per second value). As the more scattered objects you have inside Octane engine, the slower is the rendering speed. So, in the ideal case - you will get the best rendering speed if you have one big "Global" geometry object inside Octane engine.
But sometimes for very big scenes the time of translation of "Global" objects (merging all of them into one big global object during scene translation) may kill all the gain from their increased rendering speed (but for animation it will impact the first frame only, as in all other frames this mesh will not be reloaded)... So, here you just need to find some "golden mean" between translation speed and rendering speed playing with Octane geometry types...

Cool movies BTW..

[RajivMudgal]

Thanks Jim, these are great tips. We are sure to try these combinations. It would be fantastic if we could save an extra 30 to 40 sec on a frame.
Thanks Regards
RM

[nildoe]

Yeah nice movies..

JimStar.. just to clarify... i use the maxplugin of octane..so, in this case what would be the equivalent of "Global" Scatter "Movable and ReShapable in octane for max settings? i ask because in max plug we only have the option for the Movable Proxy.

Example: a scene where u have a room furnished, a skinned character, moves a "chair" from one point to another...

What setting should we put for the chair?
What setting should we put for the Skinned (deformable) character?
What setting should we put for the rest of the static objects in the scene?

Also i have been confused as to what exactly is "auto-update movable proxy"?

Nildo

[JimStar]

Yeah nice movies..

JimStar.. just to clarify... i use the maxplugin of octane..so, in this case what would be the equivalent of "Global" Scatter "Movable and ReShapable in octane for max settings? i ask because in max plug we only have the option for the Movable Proxy.

Example: a scene where u have a room furnished, a skinned character, moves a "chair" from one point to another...

What setting should we put for the chair?
What setting should we put for the Skinned (deformable) character?
What setting should we put for the rest of the static objects in the scene?

Also i have been confused as to what exactly is "auto-update movable proxy"?

Nildo

Sorry, I can only answer on that from Blender or Maya point of view. I don't know details of how exactly it should be done in Max...

In Blender or Maya your example should look like this:

• The animation mode set to "Movable proxies"
• All room furniture (except chair) should be set to "Global".
• If you have some heavy scattered not moving geometry with huge amount of instances, e.g. carpet which has a lot of scattered geometry instances - only these instanced geometry objects should be set to "Scatter" type. From the point of view of animation - this will produce static not movable geometry too, it will be loaded only in the first frame and will be just reused unchangeable in every next frame, the same as "Global" geometry. But it will eat much less GPU memory, and will take much less translation time. The less "Scatter" types you have - the faster will be the rendering itself (samples per second value).
• The chair geometry should be set to "Movable proxy". In this case - the mesh will only be loaded in the first frame, and only transform will be reloaded every next frame. It is very fast operation, so the translation takes much less time before start rendering every next frame.
• Only the character mesh should be set to "Reshapable proxy". In this case both the character's transform and the geometry will be reloaded and re-evaluated every frame.

[nildoe]

Hi Jim..

Thx for that reply..I will try to ask Karba the similarities of your terms and max terms...

Nildo

[RajivMudgal]

We have a new character in our Script. He is Kakaka the Fly. An extra terrestrial intrusion into the cartoon world.
We also wanted to test the new 2.0 Blender render engine. Wow we are loving it, and JimStar has done an excellent Job.