Gaussian Splatting has gone from being an amazing technology in the realm of research to finding its way into real-world testing and production workflows in a very short time. But there’s a world of difference between the paper, the spectacular demo, and a real-world project with clients, deadlines, and actual deliverables.

In production, it’s not enough for something to work just once under ideal conditions. There aren’t always million-dollar budgets (there almost never are!), but there’s no shortage of tight deadlines, technical limitations, legacy pipelines, and little room for improvisation. And yet, the technical and visual demands remain extremely high.

This talk offers a practical, no-nonsense look at the real-world use of 3DGS outside the lab, and how that conversation is beginning to expand toward dynamic capture as one of the most exciting avenues for the future. We’ll discuss what happens when these technologies are applied to specific projects: what works, what breaks, what can be resolved, and what is still far from being as automatic as it sometimes seems in demos.

We’ll cover the entire process from capture to the final result: acquisition systems, capturing spaces and people, volumetric video, training, visualization, real-time rendering, and integration into graphics engines. We’ll also delve into topics that are less flashy but fundamental to production: performance, formats, compression, temporal consistency, lighting, color management, ACES, HDR, and compatibility with existing pipelines.

Rather than a talk about promises, this will be a conversation about real-world use. We’ll discuss when it makes sense to use GS, when it doesn’t, what problems arise when taking it into production, and what solutions are needed so that it stops being just a spectacular technology and becomes a standard working tool.