Pixar's Universal Scene Description To Be Open-Sourced

An anonymous reader writes: Today Pixar announced their second major open source project, Universal Scene Description. USD is the technology that enables 'hundreds of artists to operate simultaneously on the same collections of assets in different contexts', says Pixar VP of software R&D, Guido Quaroni. Pixar has been working with industry to vet the new technology, gaining backing from VFX power-houses MPC and Double Negative as well as high-end digital content creation tool creator, The Foundry. Official source release is slated for summer 2016. Pixar released its RenderMan animation and rendering suite, free back in March.

Comments

It will be taken down by a DMCA request.

[the_other_one]

Pixar is too close to Pixels.

Will this fill a gap in free software?

[ciaran2014]

Does this software do something that the Blender devs currently can't do (and aren't making good progress on)? Will Blender celebrate this release?

(Note: Renderman is still proprietary, it just costs nothing for a non-commercial licence. Not free software at all.)

(Note #2: The announcement is their "intent" to release it as open source by summer 2016. Currently no code has been released and no licence can be read.)

Re:Will this fill a gap in free software?

[exomondo]

Does this software do something that the Blender devs currently can't do (and aren't making good progress on)?

This is about allowing different applications to inter-operate by having a common scene description format, it's nothing to do with free software.

(Note: Renderman is still proprietary, it just costs nothing for a non-commercial licence. Not free software at all.)

I don't think anybody implied it was Free Software, they released it free (as in the general interpretation: "without cost") for non-commercial use.

Re:Will this fill a gap in free software?

[im_thatoneguy]

If they integrated this they would have a vastly improved reference management system. However, much of what this offers probably isn't terribly useful to your average Blender user so it would be a total waste of time and actually just a time waster for someone who works by themselves and doesn't reference or instance assets or needs to move assets between multiple content creation packages.

Re:Will this fill a gap in free software?

[Carewolf]

Does this software do something that the Blender devs currently can't do (and aren't making good progress on)? Will Blender celebrate this release?

(Note: Renderman is still proprietary, it just costs nothing for a non-commercial licence. Not free software at all.)

(Note #2: The announcement is their "intent" to release it as open source by summer 2016. Currently no code has been released and no licence can be read.)

Not if it is has the same license as Renderman, which means it is only free if you don't actually use it, and don't mind signing off your soul to Pixar.

Re:Will this fill a gap in free software?

[mhenley]

Renderman is not open source. It would probably be released under a modified Apache license as per their first major piece of open source OpenSubDiv: http://graphics.pixar.com/open...

Re:Will this fill a gap in free software?

[K. S. Kyosuke]

This is about allowing different applications to inter-operate by having a common scene description format, it's nothing to do with free software.

Just like ODF is not about free software? You can't use free software (with significant penetration) if multiple software suites can't communicate (and the majority isn't using a free one).

Re:Will this fill a gap in free software?

[vmxeo]

I wouldn't be surprised if Pixar created a Blender plugin for it themselves. They recently released the Renderman for Blender plugin, so it's not out of the realm of possibility.

Re: Will this fill a gap in free software?

[exomondo]

The first sentence states it was Pixer's second open source project, so there is the suggestion...

Yes, a simple search will tell you their first one was OpenSubDiv.

Re:Will this fill a gap in free software?

[exomondo]

Just like ODF is not about free software?

Correct. It is about inter-operability between closed source software, open source software, shared source software, free software, whatever.

Nerdlinger

[Anonymous Coward]

> 'hundreds of artists to operate simultaneously on the same collections of assets in different contexts',

English, please?

Re:Nerdlinger

[Anonymous Coward]

[It allows] hundreds of mechanics to operate simultaneously on different parts of the same car.

Re:Nerdlinger

[Wescotte]

I'm thinking git for artist.

Re:Nerdlinger

[im_thatoneguy]

It's like a large application where you break up different components into separate files. Your project might include thousands of source files with hundreds of developers but since it distributes the asset information you can have one person working on a set of classes and functions while another person works on a UI that calls those classes or functions simultaneously.

In this instance you can be editing in different 'contexts' aka you can be in a 3D modeling application sculpting an asset while a lighter references that model in their set. Conversely the modeler can continue to sculpt while using the lighter's lights.

Re:Nerdlinger

[piojo]

But I gather the scene is still one file, mostly? I mean, it's easy to edit assets separately, but the summary makes it sound like almost *everything* can be edited concurrently. Have they just been really clever about how to separate every tiny little piece of data, so every detail is considered an asset? (And of course, they would need a clever way to store how information is mapped to assets, so the mappings themselves are not the cause of conflicts.)

Re:Nerdlinger

[im_thatoneguy]

The scene file could be a single file or it could be numerous files. The best comparison from my understanding of USD (and Katana's 'recipes' which clearly inspired this) is to CSS. CSS can be embedded in the document or it can be referenced, or it can reference references which reference references etc. You could have a file for a wine glass, which is referenced by a group for a place setting which is referenced by a group which is a table with place settings (but override the position of the forks to randomize the place setting slightly). The table file with the randomized place setting noise can be referenced into a dining room which is itself referenced into a house which is referenced into a block full of other houses all with referenced rooms, tables, assets and primitives.

So it could be a single file with all of that or it could be thousands of files. They say they are lazy loading too which is inspired by the Alembic file format. What this means in short is that they only load into memory the exact requested data. So you can have your camera field of view loop through the USD network and determine based off of say bounding boxes everything in view and only load each leaf that is actually meeting that filter (taking into account layering overrides etc.)

Re:Nerdlinger

[Wescotte]

Oh.. Well, in that case try and think of it as git for mechanics.

If this is what I think it is

[Applehu Akbar]

If this is really a Python-based description language for allowing animated scenes of any complexity to be described as a hierarchical set of object classes, this goes public just in time for Oculus Rift and HoloLens.

Re:COLLADA is an existing open exchange standard

[exomondo]

COLLADA is just a file format.

Re:COLLADA is an existing open exchange standard

[im_thatoneguy]

Collada doesn't to my knowledge support inheritance and overrides. So I can't have a Collada file that modifies another Collada file which modifies a third Collada file etc.

The best comparison would be CSS vs straight HTML. With CSS you can have a style sheet that modifies a second style sheet which inherits a third style sheet which overrides the property of a fourth style sheet and so on and so forth. HTML you just have a page with markup.

Re:COLLADA is an existing open exchange standard

[exomondo]

the only difference between a file format and a standard is that with a standard multiple people agree to use the same file format.

That's nice but not really relevant since USD isn't a file format or a standard.

It does provide a (both binary and ascii) file format but it is much more than that, that alone isn't particularly useful. USD itself is a scenegraph format and provides c++ libraries and python bindings for modifying that graph. You also don't have to just use usda and usdb files, in theory you could write a plugin to support any kind of file format you like.

Re:COLLADA is an existing open exchange standard

[im_thatoneguy]

But can you modify Collada references? For example if I had a collada reference to a collada file of a car can I create a collada class called "Car" that exposes functions that can be called and overridden? For instance can my car asset have a .setSpeed() function that will modify the transform matrix to move in a straight line at that velocity whenever you call car.position(10s)? (I'm not trying to be combative I'm actually curious! :D) My memory of Collada was that you could reference objects but you couldn't override a setting. So the reference was an immutable reference.

Re:COLLADA is an existing open exchange standard

[exomondo]

Yes nobody is saying COLLADA isn't good or useful but USD is obviously very different, USD isn't focussed on the file format, more the scenegraph format that it has and the libraries for manipulating it that it provides.

Re:COLLADA is an existing open exchange standard

[im_thatoneguy]

Well USD is supplanting and replacing RIB (RenderMan's scene descriptor) and ASS (Arnold's Scene descriptor). So in that regard USD *is* the providence of the rendering engine since it's supposedly the entire scene graph for the renderer.