Calendar
Slack Channel
Zoom meeting
USD on the Web prototype can be tested here.
...
Calendar
Slack Channel
Zoom meeting
USD on the Web prototype can be tested here.
Wednesday, October 4th 2023
Alan presenting project to perform animations workflows using front-end UI, submitting jobs to back-end server ultimately serving Users back with baked animation.
Revisiting productization steps of WebAssembly prototype, including actionable steps to submit a first end-to-end workflow building the lib
arch
via Emscripten from OpenUSD 23.11Based on suggestion from Nick, base branch supporting integration can be made from the soon-to-be-released OpenUSD 23.11.
Autodesk to submit Dockerfile already used for recent WebGPU prototypes, known to be working (~spring 2023)
Iterative work to be done in order to include modifications to
build-usd.py
script andcmake
files, in order to integrate Emscripten builds.(Other details available on the itemized productization list.)
Wednesday, November 2nd 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Anything interesting to show or discuss?
Proposed agenda:
High-level overview of the productization/maintenance work of the WebAssembly prototype
Wednesday, October 5th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Anything interesting to show or discuss?
Proposals for wider color gamuts on the web
Proposed agenda:
Opportunity to share list of priorities with the Google Chrome team.
Go over a state-of-the-union on WebAssembly, SIMD, WebGPU, threading, memory, etc. to establish/coordinate on priorities
Wednesday, September 7th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Special guest: Brent Scannell (Sr. Product Manager at Autodesk, Khronos glTF Working Group)
Notes:
Benefits & challenges of the vendor extensions in glTF
Contribution model intended to make it simple for vendors/implementors to prototype/share extensions to glTF
Process is rather informal, may eventually evolve to a whitepaper/RFC model to gather feedback from the community ahead of submitting extensions
Benefits & challenges of not offering a format "glTF runtime", focusing on model definition instead
Great for flexibility, getting started contributing
Can be a challenge in ensuring assets behave similarly in different implementor runtimes
Details about the glXF proposal
Intent is to bring greater interactivity to glTF experiences
Supporting scene-level interactions, as well as interaction between assets ("smart assets")
Intent is to compose scenes via referencing, using glTF assets, or other scenes
No particular prescription about the level of granularity of assets at this time.
Intent is not necessarily to dive into "simulation workflows" at this time (e.g. car collisions & deformations), but research in this space is welcome
Commonalities between USD & glTF communities
Topics about where the "authoring runtime" ends and where the "consumer runtime" beings
e.g. where would a hypothetical Unreal/Unity engine meet an embedded hypothetical USD/glTF runtime.
Challenges about asset validation, asset structure, best practices
Established test suite for validating conformity of glTF assets against expected behavior
Challenge in agreeing which renderer is the "most accurate" when conforming across vendors
glTF has rich transmission, last-mile delivery format ("The JPG of 3D assets")
USD has rich composition workflows ("The PSD of 3D assets")
Opportunities for communities to avoid solving the same problems each on their own, and meet where the formats overlap.
Invitation:
Online webinar on Delivering Interactive Experiences with glTF, September 27th, 2022
Wednesday, August 17th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Proposed agenda:
Post-SIGGRAPH 2022 regroup
Aligning with Assets Working Group regarding validation, JavaScript-to-WASM bindings for USD, etc.
Wednesday, June 15th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT) Wednesday, July 13th 13th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Note: Proposed agenda rescheduled to next meeting due to audience particularly interested in USDz validation unable to attend.
Anything interesting to show?
Proposed agenda:
From previous conversations:
USDz support can require significant effort for validating assets, and is a manual process.
Proposed action item: combine efforts to create a validation tool, if issues are difficult to raise to vendors/implementers.
Potential questions:
Have the issues resolved at the source ()
Enhance existing
usdchecker
vs. develop a different tool?Which runtime versions should be supported/specified?
etc.
Wednesday, May 18th 2022 18th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Follow-up on conversations from previous session about comparing glTF and USD for visualization
Eric Chadwick provided a description the Wayfair workflow:
Relying on 3ds Max and Maya to author and design content, using mostly VRay materials
Baking down textures and geometry to later package assets in USDz format
Strong overlap with parallel conversations about LoDs, texture/mesh compression also covered in main Working Group and Games sub-working group
Regarding interactive experiences (mobile/AR/VR/XR), fast loading time, rendering-optimized content is a must
Users/downstream consumers are unlikely to stand with their devices while waiting for content to load
Experience may be jarring if content suddenly appears in the world, or partially loads
Early conversation about combining efforts regarding asset compliance/validation for Apple's USDz format
Avoiding reinventing the wheel by sharing tools/scripts to flag potential known issues with Apple's interpretation of the format
Perhaps enhancing the
usdchecker
tool to support more options, or take inspiration from the tool and build a separate/specialized one
Wednesday, April 20th 20th 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Touch point on the principal focus of the initiative:
Common agreement that the focus of the project is not to compete or dethrone glTF as the transmission format for assets and content on the web.
glTF has specific targets, use cases and built-in optimizations specifically targeting transmission and "last-mile" delivery of assets, which is less of a concern for authoring formats such as USD.
Former user (Deleted) mentioned intent to revisit his Last mile vs. interchange article, to update the content based on latest industry development
It may be worth revisiting some of the existing documents published to state this more publicly, as this none-competitive aspect may otherwise not be clear to the audience outside the USD Working Group.
Potential future work may include publishing a "best practices" guide regarding publishing USDz assets for the Web, covering topics such as asset complexity/density, texture optimizations, known limitations, etc.
Alexander Schwank shared updates:
From the MaterialX Working Group, about Wayfair sharing their experience transitioning from a glTF pipeline towards a USD one:
see group meeting notes: April 7, 2022
From Intel, about the Sponza scene, available in glTF, USD and other formats:
Potential groups worth reaching out to regarding USD "runtime" features:
Sketchfab
Shopify
Potential Questions for the larger USD group:
Does Autodesk envision any specific use cases for WebAssembly? Any gaps, limitations with the current prototype?
jerran- While data visualization is one component, I did mention briefly in the HgiWebGPU presentation that web-based USD content authoring applications are a particular use case for this. (Note: I no longer represent Autodesk on this)
Does Pixar see any potential openings for platforms other than X86/64-bit architectures?
Wednesday, March 23rd23rd, 2022 (11:00AM PT/2:00PM ET/9:00PM GMT)
Demonstration of an early prototype of HdStorm built to WebAssembly
Demonstration of an early prototype of USD WebAssembly running within A-Frame.io :
view-filename height 250
Recent noteworthy USD WebAssembly updates:
Project proposals to bring back to the larger USD Working Group, either for standalone value or for inclusion within other projects:
Asynchronous HTTP(S) ArResolver
Wednesday, November 17, 2021 (2:00PM PT/5:00PM ET)
Wednesday, October 20, 2021 (2:00PM PT/5:00PM ET)
Low traffic
Pitch for a topic of next meeting about glTF & USD
Discussion on Materials (MaterialX, aiSurface in browser)
Update on Autodesk & legal dept.
Wednesday, September 22, 2021 (2:00PM PT/5:00PM ET)
Post discussion on <model> tag
Wednesday, August 25, 2021 (2:00PM PT/5:00PM ET)
Demo of assets viewing and interaction in Autodesk Forge in Chrome
Rendering, selection, and scene-inspection
A lot of work has been done on bindings
Custom changes to WebASM build of TBB
Wednesday, July 14, 2021 (2:00PM PT/5:00PM ET)
Conversion from USDZ into blender and out to GLTFis a great short term solution, but...
Worth doing test for conversion of animated models over time.
Makes a compelling case for a native Sdfplug-in for GLTFimport/export.
Past initial mesh import, would USDSkeland GLTFbi-directional conversion be lossless?
Draco plugin may need to support native Dracofiles better, or does it matter in a separate Sdf plugin?
KTXas native Hioplugin would likely be mandatory, does Hiosupport unpacking direct to GPU texture via Hgi? (Brian from AMD hints at a yes)
Is MaterialX and its advances (especially in tackling WebASM) going to be the way forward for all shading interchange in USD.
OCIOand color accuracy again came up as important, but too large a topic compared to getting a better bare-bones implementation anyone can use.
Wednesday, June 2, 2021 (3:00PM PT/6:00PM ET)
Autodesk showed USD import into Forge. Translates to Forge-native geometry.
Autodesk also working on a WebGPU Hydra delegate that does PBR shading.
A lot of overlap between the delegate above and Filament (i.e. SPIR-V tools running in browser)
TBB remains an issue that everyone is keen to solve.
Philippe Sawicki working on an MVP list for JS-bindings.
Dhruv Govil is interested in what can be leveraged from ASWF initiative to auto-gen Rust bindings.
Wednesday, May 5, 2021 (3:00PM PT/6:00PM ET)
JavaScript bindings, need a real strategy on how to tackle them.
...
See what changes have landed in /dev that may help reduce reliance on WebAsm-TBB and its CORS issues.
Wednesday, April 7, 2021
Working example of WebASM build of USD with a Hydra delegate for Filament
Working on macOS &
FirefoxFirefox-87.0
Ideally GitHub Pages version at version at https://marsupial.github.io/USD/extras/imaging/examples/hdFilament would be a live demo, but
due to the lack of 2 http headers WebASM using threads will fail to load
a local server.py can be used to serve the branch locally instead
Colored Quad, Sphere, and Cube are USDA. Gold monkey rotating in circle is USDZ. PBR monkey spinning around its center is glTF.
HdFilament shader generation is minimal and displayColor is used for primitives. Textures on monkey.usdz are not loading correctly.
USDA loaded can be edited in the HTML: https://github.com/marsupial/USD/blob/HdFilament/extras/imaging/examples/hdFilament/index.html#L119
Wednesday, March 10, 2021
- Roman Zulak
- Philippe Sawicki
- Alexander Schwank
- Dhruv Govil
- Alex Gerveshi
- Nick Porcino
- JT Nelson
Goal is to have USD WebAsm branch and a WebAsm build of Filament Hydra Delegate by next meeting (April 7).
Gain a better understanding of performance and limitations loading arbitrary assets and materials.
Let a larger audience see the size of changes needed to support the platform.
If the changes are too large for Pixar to mainline, what options are there?
Provide a use case for Filament that could help push HdFilament code into Google's repo.
Initially may be a private repo/branch not totally accessible to the world.
JavaScript bindings to USD.
Roman has started porting a ui-handle implementation to WebGL that would be simplified with re-use of Gf/math library.
What is the cost (in terms of source code) for additional bindings.
Python support via browser (à la Pyodide)?
Is there any gain in having existing Python scripts run in a browser?
May be too expensive of a runtime environment.
Plugins and external dependencies for a WebASM build:
http(s) ArResolver would be very useful once everything starts working.
OpenSubdiv is probably a good candidate for a next-step.
Is OpenVDB support in WebASM via NanoVDB a possibility (or useful)?
Qt WebASM is interesting, but a little too unstable right now.