Open Source Days 2026

Founded in 2018 by the Academy of Motion Picture Arts and Sciences and The Linux Foundation, the Academy Software Foundation (ASWF) is the motion picture industry’s premier organization for advancing open source software development.

David Morin, Executive Director of the Academy Software Foundation, will share updates and milestones from the past year along with a look ahead and what's to come.

Explore the evolution of MoonRay as it transitions from a proprietary DreamWorks tool to a cornerstone of open-source rendering.  We’ll cover the technical challenges of "breaking out" of the studio, the strategy and lessons learned on building a sustainable community, and why the Academy Software Foundation is the vital next chapter for the project and the path forward.

Project updates (and some TAC stuff) given by the TAC chair.

We now have a lot of projects, and likely folks that don't know about many of them. This would be a chance to highlight cool work on the main stage, and promote BoFs the following day / town halls.

OpenVDB was the first project accepted into the ASWF in 2018 and has continued to establish itself as the ubiquitous choice for sparse volumes in VFX and Animation. However, the file format underlying it has remained unchanged since 2014. The newer tools and techniques available today make a strong case for modernization.

This session will provide an overview of some of the key open-source file formats in use in the VFX and Animation industry today, and discuss the approaches they have taken to I/O and compression, before looking at some of the latest libraries and techniques that have become widely adopted. We will do a deep dive into the new OpenVDB file format in which these lessons directly informed the design, using performance and compression data to provide clear motivation for a breaking change. Finally, we will share details of our new extensible codec framework and invite the audience to contribute new ideas and develop new codecs to improve on those shipped with OpenVDB.

In addition to the significant changes to our file format, we will also briefly go over major changes—especially to NanoVDB—which now supports many novel dynamic (vs. static) applications of sparse volumes on the GPU. This includes constructing grids from meshes and points, deforming level set surfaces, merging and pruning grids on the GPU, as well as fundamental topology operations like morphological dilation. Finally, we will also talk about new changes to our toolset, in particular vectorization and vdb_tool.

Open-source DCCs ship a generic core; studios need something calibrated to their pipeline, their shows, and their dailies room. We argue that gap should be closed by small teams shipping in days, not quarters — and that AI-augmented engineering finally makes that realistic without devolving into vibe coding.

When Rodeo FX adopted xStudio, the upstream release was the foundation. In months, a small team built the rest: a forked C++ media hook for OCIO and cross-platform paths, plus 14+ Python plugins — a ShotGrid browser, auto-conform from track names, version compare, notes publishing with annotations, dailies broadcast, HUD overlays, snippets.

This talk shares the methodology behind shipping production-grade xStudio extensions on top of an open-source core at studio cadence: what the plugin architecture made easy, where it bit us, and which patterns we want to push back upstream so the next adopter doesn't rebuild the scaffolding.

This session presents two color-management case studies from Japanese 2D animation, based on Tomohiro’s work as a post-production engineer collaborating with production teams and introducing OpenColorIO workflows. The production case discusses a tailored OpenColorIO config and HDR/SDR tone mapping developed after evaluating ACES configs and ACES RRT. It focuses on preserving painted artwork colors while extending digitally created light and optical effects in HDR, and sharing SDR/HDR views between production and post-production. The remastering case discusses ADX encoding of film scans and LMT-compatible look design informed by print timing values, CMY dye transmittance, and projection light spectra. It focuses on reproducing color close to the original print while adapting a print-derived look for HDR delivery. The session shares reusable requirements and design patterns for OpenColorIO configs, HDR/SDR view design, ACES-aligned Look Transform design, and validation.

Two decades ago Van Gogh’s painting of Two Sunflowers played a key role in shaping best practice for artwork reproduction that ultimately led to international standards. These standards defined scene-referred imaging validation techniques used worldwide to this day for artwork reproduction.

In July 2025 the Metropolitan Museum of Art Imaging team set out to create new 2D and 3D images of this same painting to explore objective color validation techniques for 3D scenes. The project was an opportunity to invite experts from the AOUSD community to be onsite at the Met studios for this important capture session to gain insight into the cultural heritage imaging community.

In this presentation, Scott Geffert will share the results of this special imaging session and the efforts underway to develop open tools to improve color fidelity for 3D renditions and online representations.

Lens metadata should connect what happened on set to what happens in post, but in many productions it is incomplete, mistrusted or lost before it reaches VFX. In this 10-minute lightning talk, Gary Keller, General Manager of DCS, will introduce an industry survey designed to measure where lens metadata breaks down, how often it affects matchmove, layout, compositing and virtual production workflows, and what developers need from camera, lens, metadata and pipeline vendors. The talk will explain why DCS is running the survey, how the anonymised findings will be shared openly, and how the data can support better open-source tooling, interoperability and standards discussions. Attendees will leave with a clearer view of the problem, an invitation to contribute, and a practical dataset they can use to inform future pipeline work.

Because of new developments in realtime game engines and large LED volumes, in-camera visual effects (ICVFX) has become a common form of virtual production used by cinematographers and filmmakers worldwide. Yet there is no Standard Evaluation Material for ICVFX available to filmmakers.

Developed by the Motion Imaging Technology Council (MITEC) of the American Society of Cinematographers (ASC), the Standard Evaluation Material 3 project for virtual production (StEM3-VP) will provide freely and readily available content on the ASWF Digital Production Example Library (DPEL) to help filmmakers evaluate in-camera visual effects stages.

Unreal Engine 5.8 introduces the new Tiled Mipmap Video format, which delivers high-performance in-engine video playback with tiled streaming, smooth scrubbing, HDR support, and efficient compression, reducing the storage and bandwidth demands of high-resolution EXR workflows for ICVFX and live event production.

TMV is an extension to the OpenAPV codec and this talk details Epic Games' work to design and build the feature as well as its collaboration with the Academy Software Foundation to contribute it to the open source community.

Security, trust, and code safety is a top priority for VFX, animation, and production studios. New AI resources are coming to market daily, but adoption can’t come at the risk of safety and security. While solutions built on open source technology have clear benefits, from reducing operational costs to simplified licensing, customizations and modifications without the right guardrails may introduce risks. Implemented with the right protections, open source AI orchestration is allowing studios and independent creators alike to accelerate and unlock new workflows with the latest AI models while maintaining control and security. Additionally, when Apache 2.0 is used, as opposed to other license options, studios have more agency over development and deeper control over pipeline access. This session will cover key considerations, benefits and practical tips for integrating AI within production-grade environments using open source.

Industry standards like MaterialX, USD, and OpenPBR are key to enabling collaboration between individual artists and across production pipelines that use different DCC applications. OpenPBR is a unified shading model that allows artists to design and share assets between different applications while preserving their appearance.
In this talk, we will present insights into Blender’s journey toward adopting OpenPBR support. We will discuss the challenges we faced during our integration, including discrepancies between our OSL backend, which used code generated by MaterialX’s OSL code generator, and our strictly spec-based implementation in Cycles’ own SVM shading system.
To analyze these discrepancies, we collaborated with the OpenPBR team and began developing a testing framework to compare the appearance of different OpenPBR material setups across renderers (e.g., Cycles, Arnold, or Adobe's OpenPBR library). The framework will improve consistency in the future, when using OpenPBR.

his session explores the growing convergence of OpenUSD, MaterialX, and OpenPBR as a unified open-source rendering pipeline for VFX and animation production. Speaker presents the latest developments in how MaterialX shader networks and OpenPBR material definitions are being authored, exchanged, and rendered within USD-based pipelines. The session covers practical integration patterns, interoperability challenges solved in recent open-source releases, and a demonstration of an open-source reference renderer that exercises the full USD + MaterialX + OpenPBR stack. Attendees will leave with a concrete understanding of the current state of physically based shading in open-source pipelines and actionable guidance for adopting these standards in their own studio workflows.

In this presentation we will talk about the cross Lucasfilm/Pixar collaboration to bring arbitrary MaterialX BSDF networks to RenderMan XPU. We'll highlight the approach we took to allow the same source code to run seamlessly on CPU and GPUs -- opening the door to MaterialX Standard Surface and OpenPBR. We will use a new asset we will make available to the community as the basis for our talk.

In order to achieve the dream of reliable, consistent results in lighting and materials across OpenUSD applications, we need a reference implementation and compliance test suite for the various emerging standards across such projects as UsdLux and MaterialX for developers and users to refer to.

This talk describes an open-source, cross-platform, reference path tracer implemented as an extension to the OpenUSD hdEmbree delegate. Developed by engineers at NVIDIA, this path tracer builds on the rich ecosystem of open-source foundations such as MaterialX and OpenQMC to provide a framework for the community to share knowledge and development of rendering standards that will enable seamless rendering interoperability.

One of the primary challenges with OpenUSD proposals is aligning on the problem statement that cross-domain communities agree on. OpenUSD’s reach extends beyond film and animation, into gaming, robotics, simulation, and geospatial analytics. That growth strains the proposal process - the community is increasingly diverse with unique, domain specific requirements and procedures. Groups new to USD, with their own vocabulary and experience must agree not just on solutions, but first a common understanding must be built on what problem is being solved.

This session covers the methodology used across many proposals to meet this challenge. A clear, agreed-upon problem statement, not a solution, is the first gate– and the real work there is disentangling terminology collisions, and stepping back from USD itself to state a concern in neutral terms.

We’ll show how this serves both cross-cutting concerns (Source Identifiers, Authorship, IP Protection, and Profiles) and domain-specific features like LOD, Image Planes, BReps, and Geospatial, with Profiles the key mechanism for bucketing discussions, audiences, and priorities.

From this presentation, attendees leave knowing how to read a proposal, write a problem statement the community can reason about, and engage through proposal PRs, prototypes, and AOUSD.

BowerBot started as a way to import assets from different
sources like Sketchfab and place them into OpenUSD scenes
organized the way production pipelines expect, so artists
could focus on refining layout composition instead of fixing
layer structure. The project has since grown into an Apache
2.0 open-source AI agent that authors full OpenUSD scenes
from natural language while following guidelines from the
Academy Software Foundation, Pixar, NVIDIA, and the Alliance
for OpenUSD. Unlike AI tools that generate Python code on the
fly, BowerBot uses a tool-based architecture. The agent calls
a fixed set of pre-built tools that wrap the pxr Python
bindings, with ASWF guidelines enforced inside the tool
implementations.
BowerBot's plugin system lets studios connect
their own asset managers and pipeline tools as skills. The
session covers the architecture, the patterns it enforces,
and why tool-based agents produce more reliable USD than
code-generating alternatives.