Multiverse | Studio: with Alembic, USD, MaterialX and unseen performance in Maya

[p8erto]

Hello there, long time no see dear Maya user community!

We just released a trial of Multiverse | Studio v4 -- with Alembic, USD, MaterialX support in Maya.

This release introduces some amazing features and unseen performance improvements.

Read along. Text includes links to videos.

Website: http://multi-verse.io

Vimeo: https://vimeo.com/jcubeinc

What's New
=========

Full/Detailed release notes are here:

  https://paper.dropbox.com/doc/Multiverse-Studio-Release-Notes-cssH92yDmHRCQa7tu8ayJ

- Ultra fast and memory efficient compared to Maya built-in tools. In the following videos we are talking 10-20x, but in production we have seen 100x and ability to handle data that _cannot_ be handled (therefore unmeasurable):

  - animated asset, complex hierarchy -- https://vimeo.com/215624720

  - static asset, heavy geo -- https://vimeo.com/215624720

- Unified Asset Write/Read architecture using the new "Hyperspace" API
which wraps Alembic & USD and works on any platform / architecture.
Note that Alembic is accessed directly and not behind the USD Sdf
interface and is therefore at its ideal performance.

- Completely unified UI: now you can read/write Assets regardless of
their data format or backend, this is why we call them just "Assets".

- Sample interpolation in Maya VP2 and during procedural rendering
both in Maya and KATANA. This means you get free (linear) 3D motion
blur even with just 1 sample per frame (space saving of 50% in case of
deformation motion blur) -- https://vimeo.com/213962871

- Timing & re-timing controls for Assets (very useful for crowds) --
https://vimeo.com/214436067

- Complete primitives and property support: poly, subd, points,
curves, normals, tangents, creases, visibility etc...

- Material exchange of full shading networks, including displacement
via MaterialX (see MaterialXS open source project --
https://github.com/j-cube/materialxs). Works from Maya to Katana using
3Delight for Katana -- https://vimeo.com/215588519

- High performance Asset Read with "Elastic Compounds", can be
unpacked, packed, and drawn in the VP2 fast and with super efficient
memory use -- https://vimeo.com/214434209

- Improved bounding box reading mode now can read any Asset regardless
of inherent hierarchical complexity.

- Now includes a full distribution of USD, including Python bindings,
on all platforms and on all Maya versions.

- Multi-task writing: (scripted) ability to write multiple Assets in
one go with one Maya DAG evaluation --
http://multi-verse.io/plugins/maya/tools/#assetwrite

- Support of several HOUDINI features: velocity motion blur, packed
instances etc. --https://paper.dropbox.com/doc/Houdini-User-Reference-Guidelines-ZBBZcRQvRUwXtoPcKYIPf

- Largely improved 3Delight procedural

- Improved PRMan procedural (subdivs, creases, UVs, motion blur...)

- Arnold 5 support is WIP, working also on adding VRay soon™ 😉

- New "Elastic Compositions": ability to write and read USD
compositions in Maya (we call them elastic because once read you can
unpack Assets) -- Currently we recommend using USD files since Alembic
ones are read through there SDF API and performs poorly--
http://multi-verse.io/plugins/maya/tools/#compwrite -- We plan to
overcome this in the future.

- New high level documentation
(http://multi-verse.io/plugins/maya/tools), regular documentation,
release notes & website.

[Anonymous]

There's a lot in this one, great job guys! I'm a single person, but I'm willing to guess your userbase will be very pleased with this release. That frame interpolation looks killer, can't wait to use Multiverse one day. All I need is Redshift support 😉

Mike

[p8erto]

A couple of videos about USD support:

1. Writing & Reading Multiverse "Elastic Composition" via USD -- https://vimeo.com/215930511
2. Native Alembic / USD writing, reading, VP2 drawing and rendering with motion blur via sample interpolation -- https://vimeo.com/215937658

[p8erto]

Regarding redshift support, you have to consider that you must be able to feed the renderer with multiple Alembic files which have up to 100GB of data so I am not sure how well it would handle that, memory wise 🙂

Furthermore there is no public sdk for their rsProcedurals yet, so until that does not happen it will be unlikely for us to support it. If a paying customer requests that we will look into it, of course. It would be also fun to compare it with offline renderers which of course have zero problems memory wise.

[Anonymous]

Oh, I'm not implying anything soon, or even in the next few years 😉 I know how long developing stuff like this takes, so no rush on my part. I appreciate the work you guys have done, and the new USD work will definitely play a huge part for DCC users over the next few years.

The Redshift public SDK is something we're hoping to see by the end of this year, I think they're waiting until the shading packages are unified. The memory thing is definitely the crux of the issue with something like this. Redshift's out-of-core system would technically allow something like this to work, but you'd lose a lot of the speed benefit of pure GPU rendering. I'm really looking forward to GPU's carrying massive amounts of VRAM, and I'd also be curious how technologies like NVIDIA's NV-Link will play into this.

[p8erto]

> Oh, I'm not implying anything soon, or even in the next few years 😉 I know how long developing stuff like this takes, so no rush on my part.

Actually, thanks to our architecture adding a renderer's support is not so much work. The biggest problem with a renderer like redshift is its dependency on hardware that would overcomplicate by a large factor the testing / QA process introducing more manual labor and other variables.

> I appreciate the work you guys have done, and the new USD work will definitely play a huge part for DCC users over the next few years.

Thanks!

> The Redshift public SDK is something we're hoping to see by the end of this year, I think they're waiting until the shading packages are unified. The memory thing is definitely the crux of the issue with something like this. Redshift's out-of-core system would technically allow something like this to work, but you'd lose a lot of the speed benefit of pure GPU rendering.

This is precisely the issue with it. You should not rule out offline rendering, things are fast rand faster and when super complex scenes come into play you see the advantages allover 😉

[p8erto]

Hey Mike,

just FYI: we are adding support for RedShift via rsProcedurals. Feel free to spread the news on their forum 😉

 Cheers.

[Anonymous]

Thanks for the info update! I'll be sure to let everyone know 😉

[jch]

Hey Pablo,

Any ETA on Vray support? We are evaluation Katana with Vray. Although I am still wrapping my head around Katana and Vray, I have a vibe based on videos you have posted online that Multivserse may facilitate the process of bring lookdeved assets from Maya to Katana for setting up lighting and render layers.

[p8erto]

> Any ETA on Vray support? We are evaluation Katana with Vray. Although I am still wrapping my head around Katana and Vray, I have a vibe based on videos you have posted online that Multivserse may facilitate the process of bring lookdeved assets from Maya to Katana for setting up lighting and render layers.

Absolutely, there are huge benefits in both Maya and Katana.

About VRay support: if you are a VRay client please send a mail to Vlado / VRay support so the priority gets bumped up.

P

[p8erto]

Hello folks,

pre SIGGRAPH updates!

In this video you can see MaterialX transfer with Multiverse | Studio v4.4.1 to be released later today (http://multi-verse.io) with Foundry KATANA and Autodesk Maya using the Arnold renderer.

[video]

You can write/read/use of full shading networks (including displacement) with Arnold 5 in both a "Maya to Maya" scenario and a "Maya to Katana" scenario. Full shading networks transfer is powered by J CUBE's MaterialXS, an "eXtra Small" Python implementation of MaterialX (https://github.com/j-cube/materialxs)

Also for who did not hear the news, Multiverse | Studio is now distributed by Foundry (press release: announcement).

Enjoy!

P