Sebastien Speierer

Less than two months to the EGSR papers deadline! If you have some cool research on rendering and synthetic image generation in general, submit it to EGSR and join us in Bordeaux (France) next July! Details: egsr2026.inria.fr

In the last months, @seb_spe and I have been working on generalizing Gaussian splatting to non-exponential image formation models. TL; DR: Using faster-than-exponential transmittance we can achieve up to 4x faster rendering without severe quality loss.

@janusch_patas @RobinerdGames The code released here can do this already! And it is easy to build on top of it to implement the other approaches 😉

@RobinerdGames There are already approaches that are doing ray tracing. However, none of the more promising ones has released code. If that's not coming, I should implement it at one point myself :)

Don't Splat Your Gaussians: Volumetric Ray-Traced Primitives for Modeling and Rendering Scattering and Emissive Media Not a new paper, but the code is released (MIT license) today! Contributions: • A novel kernel-based primitive representation for volumetric media that fits into the radiative transfer framework and can be integrated into any physics-based rendering engine. • Closed-form expressions for transmittance and emission, as well as distance sampling routines proportional to transmittance, for 3D Gaussian and Epanechnikov kernels. • An efficient ray tracing-based implementation for solving light transport using our novel volumetric representation for both scattering and emissive (radiance field) media. • Adjoint derivatives of our forward methods for solving inverse reconstruction problems efficiently.

Very excited to share our latest work on Volumetric Primitives which will be presented at SIGGRAPH25 🚀

Following over 1.5 years of hard work (w/@njroussel& Rami Tabbara), we just released a brand-new version of Dr.Jit (v1.0), my lab's differentiable rendering compiler along with an updated Mitsuba (v3.6). The list of changes is insanely long—here is what we're most excited about🧵

"Don’t Splat your Gaussians: Volumetric Ray-Traced Primitives for Modeling and Rendering Scattering and Emissive Media" arxiv.org/abs/2405.15425

@jon_barron Exciting work! It would have been great to see our paper cited directly in the publication as well. 😉

Many other people in the community have been pursuing this "what if 3DGS but raytracing" angle, here are some other great recent papers in this space that are worth checking out: gaussiantracer.github.io raygauss.github.io arxiv.org/abs/2406.09733 arxiv.org/abs/2405.15425

We dropped a great paper today! Frustrated by some of the limits of Gaussian Splatting, we developed EVER: a drop-in 3DGS replacement that raytraces constant-density ellipsoids. This gives much higher image quality and flexibility than 3DGS while retaining real-time speeds.

@jon_barron Sharing this here in case someone is interested to learn how Epanechnikov kernels can be used as a replacement to 3D Gaussians to improve the rendering performances 🚀 arxiv.org/abs/2405.15425

What is the truncated quadratic probability distribution `3/(4*c)*max(0, 1-(x/c)^2)` called? It's such a basic thing but I can't find a name for it anywhere in the literature.

@wenzeljakob Congratulations Wenzel! 🥳

I was promoted to associate professor 🥳

The free web version of "Physically Based Rendering: From Theory To Practice" is now based on the 4th edition of the book. Enjoy! (Link: pbr-book.org)

Excited to hear our Open Source Days keynote speaker, Wenzel Jakob, talk about Strategies for Open Source in Academia! #aswf #opensource

Kudos to the amazing artists and engineers at Wētā FX! This is a true masterpiece, visually stunning, and I am really proud to have worked on this movie during my time in New Zealand. Can’t wait to go watch it a second time in the theatre! #AvatarTheWayOfWater

Job openings at the Realistic Graphics Lab! If you are interested in working on Mitsuba 3, Dr.Jit, or doing research in differentiable rendering, this is the place to be! I can't recommend it enough! 🇨🇭

After working for +3 years on Mitsuba 3, I am now ready to move forward in my career and bring physically-based inverse rendering to the industry. So if you and your team are interested in such topics, feel free to reach out. Or let’s chat in person at SIGGRAPH next week! 🚀

We've released the implementation for our paper on unbiased inverse volume rendering: github.com/rgl-epfl/unbia… It uses Mitsuba 3 and is implemented as an Integrator plugin. There's a script to easily reproduce the inverse reconstructions from the paper.

With the recent release of the new Mitsuba 3 system I can now finally also release the code of our "Monte Carlo Estimators for Differential Light Transport" SIGGRAPH paper from last year: github.com/tizian/differe… (Joint work with @seb_spe, @iliyang, and @wenzeljakob.)

Holy smokes @tizianzeltner's writeup about polarization in the Mitsuba 3 docs is incredible: mitsuba.readthedocs.io/en/latest/src/…

We've just released the implementation of our Siggraph 2022 paper on "Differentiable Signed Distance Function Rendering" on Github: github.com/rgl-epfl/diffe…. The code allows to optimize SDFs from (synthetic) reference images and is based on Mitsuba 3/Dr.Jit! (1/3)

Mitsuba 3 is out! 🚀 This has been a wild journey but I am really excited and proud to finally make this framework available to all graphic researchers & engineers around the world. Can’t wait to see what you will do with it!