|
Back to List
Production 2026-05-06

NMIXX 'CRESCENDO' MV: From On-set Data Acquisition to 3DGS Web Viewer

Innovating VFX Pipelines & Communication through LiDAR Scanning, Marking Setup, and Gaussian Splatting

Written by 75mm Studio

Recently, 75mm Studio participated in the new music video project 'CRESCENDO' for NMIXX. We successfully carried out the entire workflow—from acquiring on-set data, which serves as the core foundation for visual effects (VFX), to creating 3D Gaussian Splatting (3DGS) assets, the latest 3D reconstruction technology.

In this post, we share the flow of our pipeline: how a physical set was transformed into digital data to maximize efficiency, and how this data expanded into an innovative tool for inter-departmental communication as well as a web-based interactive experience.

1. Laying the Invisible Foundation: Marking Setup and LiDAR Scanning

For seamless CG compositing, it is essential to bring on-set camera movements and spatial information into the digital domain without error. To achieve this, the 75mm Studio on-set team placed precise tracking markers throughout the set prior to shooting. These markers serve as anchor points for camera matchmoving in post-production.

Simultaneously with marking, we conducted a full LiDAR scan of the set. By emitting millions of laser pulses and measuring their return times to record the environment as a 3D point cloud, this process snapshots the physical scale and structure of the set with millimeter accuracy.

LiDAR scan 1LiDAR scan 2

2. Maximizing Project Efficiency: Interconnectivity and Parallel Processing

In traditional workflows, full-scale CG production often had to wait until filming ended and offline rough cuts were finalized. However, on-set LiDAR scan data enables true 'Parallel Processing'.

Because accurate 3D spatial data (digital twin) of the set is handed over to post-production immediately after shooting, the CG team can configure layouts and test camera paths in advance without waiting for final edits.

On-set Data Acquisition ➔ Matchmove ➔ CG Asset Placement are organically linked, significantly shortening the overall project timeline while raising the quality of final deliverables.

Matchmove 1Matchmove 2Matchmove 3Matchmove 4

3. Gaussian Splatting (3DGS) as an Innovative Communication Tool

In this project, our major focus was introducing 3D Gaussian Splatting (3DGS). While LiDAR data guarantees spatial 'accuracy', 3DGS reconstructs lighting, textures, and ambiance into photorealistic 3D representations.

Technical Details & FAQ

What is LiDAR scanning and how is it used in VFX?
LiDAR (Light Detection and Ranging) is an ultra-precise 3D scanning technology that uses laser pulses to measure spatial distances. In modern VFX pipelines, precise geometry data acquired through LiDAR ensures that computer-generated CGI environments match physical movie sets with zero margin of error. This maximizes the accuracy of camera matchmoving and virtual set extensions, drastically reducing post-production turnaround times.
How does 75mm Studio process massive Point Cloud data?
Raw point cloud data acquired on-site consists of billions of points, making it too heavy for standard DCC software like Maya or Unreal Engine. We run the raw data through an advanced proprietary processing pipeline that involves point registration, aggressive noise reduction, and topological lightweighting. Ultimately, the dense point cloud is converted via meshing and retopology into an optimized, highly manageable polygonal asset.
What is the primary difference between Photogrammetry and LiDAR?
While LiDAR structural scanning focuses on calculating exact physical dimensions with millimeter accuracy (the 'skeleton'), Photogrammetry focuses strictly on generating photorealistic visual textures (the 'skin'). At 75mm Studio, we employ a hybrid workflow combining the undeniable millimeter-accuracy of laser scanning with the hyper-realistic optical fidelity of photogrammetry.
Why is 3D Gaussian Splatting revolutionary for spatial archiving?
3D Gaussian Splatting is a breakthrough AI-based neural rendering technique that bypasses the limitations of traditional polygonal rendering. It easily captures incredibly complex micro-details such as translucent objects, fine foliage, and specularity. Moreover, it achieves this extreme photorealism while maintaining real-time 60fps render performance, even directly within a web browser, making it the ultimate game-changer for digital twin distribution.

Interested in this technology? Contact Us