Summarizing my Siggraph 2025

Siggraph 2025 in Vancouver, Canada has wrapped up a couple weeks ago. I’ve been lucky enough to attend twice now because of my role at IKEA and both times have been excellent. This time around I thought I would put into writing some of the learnings I bring with me and what I intend to do next with them.

DigiPro

While DigiPro isn’t officially part of Siggraph the fact that its co-located allowed me to attend. With its single track and focus on hands-on information it serves as a great opener to Siggraph.

One of the first talks were on OpenPBR, a reoccurring topic for the rest of Siggraph. OpenPBR seems to be quickly establishing itself as the de-facto standard for PBR surfaces. The most exciting news in my opinion was the incoming real-time implementation. There is currently already a working real-time OpenPBR renderer in Substance Designer and there are plans on either creating a specification or reference implementation. I’m sure this could have a significant impact on how we render physical materials across multiple real-time domains in the future, especially where we use GLTF today.

Another interesting topic was Netflix Animation Studios showcasing a tool dubbed “Launcher”. This tool creates what they call “hybrid runtime environments”. Combining traditional dynamic package management (using Rez) together with traditional containers allow them to easily manage and scale both end-user and compute environments. Simple in theory, no doubt very challenging execution.

The team behind Predator: Killer of Killers was present to showcase the importance and power of real-time tooling for modern VFX productions. One example was Hummingbird, their custom compositing tool in Unreal Engine, that allowed an artist to light and compose a shot in less than a day, truly highlighting the importance of real-time feedback for artists.

Katana was present to talk about what they call “USDSuperLayer”. This new construct is focused on managing large numbers of primitives or large shading graphs represented as nodes in USD. An example would be the handling of thousands of similar lights in a scene where this new construct would help in doing so. While I am sure this is very useful in Katana I am generally a bit wary of tooling specific functionality when it comes to cross platform technologies such as USD but I am looking forward to trying it out.

ASWF Open Source Days

With the growing importance of open source attending the ASWF Open Source days at Siggraph feels like a no-brainer. This year had a good mix of topics ranging from how studios have leveraged open source software for success to showcasing new interesting projects.

Two of the most interesting talks where Security Audits are Not Scary - Applying a Novel Approach to Improving Open Source and My Boss Needs To Know: Open Source Is the Future (Help Me Prove It!). The former highlights some of the efforts going on in the open source space to strengthen security which has obviously become highly relevant recently. Having the greater community take ownership of these security audits should hopefully simplify adoption of open source libraries especially in highly controlled enterprise environments. The latter will be very familiar to anyone who has tried to drive open source adoption, both using it and contributing! It’s also a great speed run in talking the language of other roles and what matters to people who might not be developers.

Some other highlights include Flowpipe, SYCL and ShadingLanguageX. All very interesting open source projects that I believe will have many practical applications going forward.

All of the talks are available on Youtube. I highly recommend brewing a pot of coffee and setting aside the time to go through them all.

Siggraph 2025

Obviously Siggraph is huge and I won’t cover it all in detail but let’s look at some of my highlights and overarching themes I encountered. Because of my work at IKEA my primary focus was on USD, MaterialX and VFX pipeline development. While these topics did not dominate to the same extent as last year there was still a great selection of talks to attend.

OpenUSD

Practical discussions around USD still seem very focused on how to develop pipelines, asset resolution and conformance across 3D digital content creation (DCC) tools and renderers.

A notable talk was on the foundations of production using USD in your VFX pipeline by Nick Porcino. One of the more important takeaways, which might seem obvious if you’re an experienced software developer, was the impact of building ad-hoc pipelines. The ad-hoc pipeline, growing organically as requirements come in, is ever prevalent in all aspects of software development but VFX is probably especially vulnerable with pressured timelines and new artistic visions over the course of running productions. Eventually you end up with something brittle that’s hard to extend and has significant people dependencies. This might actually have been more acceptable back in the day when you were largely tied to a very locked down DCC ecosystem but with USD this is no longer the case. The portability between DCCs means you want to be able to pivot depending on the demands of your productions. So think about your pipeline fundamentals and compartmentalize functionality into modules, perhaps across boundaries defined by frequency of change?

Asset resolution is an ever present problem and I think the main takeaway here is the same as above. Think about your USD structure in detail and make sure you have the procedures in place to change the structure if necessary, which it is almost guaranteed to be. You don’t want your entire ecosystem to break when that happens. Pinning versions, URI schemas, cache optimizations, enough material to fill a lifetime here.

A real highlight was the new USDLux specification. Previously USD did not clearly detail how to interpret lights which resulted in wildly different results depending on your renderer.

Karma

Arnold

This new specification details how to properly create a conformant renderer, something which is already being adopted by vendors. As a bonus this serves as guide for anyone wanting to implement their own Hydra delegate. Together with the new physical camera specification conformance across renderers is looking promising.

MaterialX

The talks I attended on MaterialX where largely focused on OpenPBR and to some extent the merits of translation graphs for GLTF conversions. There’s not a lot going on here that you couldn’t just find by reading online but if you haven’t looked at OpenPBR yet I encourage you to do so. I think it’s a very good development for the industry as a whole. If you’re curious about the work we do at IKEA using translation graphs I recommend this talk by my colleague Johan Bowald from this year’s MaterialX Virtual town hall.

Real-time graphics and Vulkan

I was fortunate enough to have some gaps in my schedule that allowed me to attend a few courses on real-time rendering and Vulkan. This isn’t currently part of my work but if you’ve read some of my previous blog entries you’ll know it’s one of my passions.

GPU work graphs

First off was GPU work graphs, an amazing new development allowing the GPU to dispatch shader invocations without a CPU round trip. The course provided was comprehensive with live coding, robust samples and some more advanced projects to bring home as well. I’m currently waiting to see what comes of Slang’s work graph implementation before looking at adoption in my current toy rendering project.

Advances in Real-Time Rendering

For those who don’t know, the Advances in Real-Time Rendering course has been around for a mind blowing 20 years at this point. Several game companies go into incredible detail on their rendering implementations and its a beautiful thing. Highlights here included a walkthrough on global illumination in idTech8 and mobile rendering at HypeHype. It’s all very crunchy but to me, even as an outsider to the games industry at the moment, it still feels very valuable to be able to follow these kind of courses and insights for my personal toy projects.

Modern Vulkan course

Staying on top of new development in aging technologies is always a challenge which is why I really appreciated the course on modern Vulkan this year. Most of it validated research I’ve done on my own with a focus on dynamic rendering, bindless resources and the “new” synchronization primitives (timeline semaphores). The combination of dynamic rendering together with the new unified image layouts should make a world of difference for renderpass handling. If you’re into Vulkan but haven’t looked at new developments in a while I highly recommend checking out the above topics including dynamic rendering local reads for tiled GPUs.

Blender community meetup

This is really just a side note but it was really amazing to experience the passion exhibited at the Blender community meetup. Even with a packed room we all got the opportunity to introduce ourselves, opening up for making new connections in the community. It let me meet a ton of great people and have lots of interesting discussions.

What next?

Well to summarize I had an amazing time in Vancouver this year. This short blog post only covers a fraction of what I encountered and the conversations that where had. I was especially happy to be able to spend more time with my colleagues, getting to know them more and learning new perspectives.

Up next I am mostly excited about applying some new ideas for our USD pipelines and workflows at work. There’s plenty to explore in this space and I hope we will be able to talk about it publicly soon. Other than that I am looking forward to applying some of the modern Vulkan concepts to my ongoing toy renderer project.

As always, thanks for reading. Leave a comment if you were at Siggraph or if you feel like I missed something from my experience you’d like know more about.