Play Dirty

Play Dirty is Shane Black’s 2025 film for Prime Video starring Mark Wahlberg, grounding its gritty, high-stakes world in a visually immersive reality. The film blends practical action sequences with digital enhancements — complex stunt work, set extensions and detailed digital environments — without compromising the authenticity of its real-world locations.

SCANable played a pivotal role in bringing that world to life. Leveraging advanced 3D scanning, photogrammetry and LiDAR technologies, the team captured real-world environments and props with cinematic precision, from accurate digital doubles to fully reconstructed locations. These point-cloud and photogrammetric assets gave the filmmakers and VFX teams the flexibility to push visual storytelling while keeping the film’s gritty tone firmly rooted in realism.

Matchmove needs the room, not the plate

Camera tracking software solves for a camera by finding features in the image and inferring how they moved. On clean, textured, well-lit footage that works well.

Action footage is rarely any of those things. Handheld camera, motion blur, smoke, shallow depth of field, and long lenses on flat surfaces strip out exactly the features a solver depends on. The solve becomes unstable, or it converges on a camera path that is mathematically plausible and physically wrong.

A point cloud of the actual location changes the problem. Instead of inferring the geometry and the camera at once, a matchmove artist already knows the geometry and only has to solve for the camera. Tracks lock down faster and hold under shots that would otherwise defeat them. That is one of the main reasons set and location scanning happens on action productions.

Where a mesh stops describing the subject

Polygons describe surfaces. They are an excellent model for a wall, a vehicle, or a body, and a poor one for anything without a definable surface.

Foliage, chain-link, wet reflective ground, scaffolding, and heavily layered urban clutter all defeat conventional meshing — the geometry either explodes in complexity or smooths away the detail that made the subject look like itself. Photogrammetry can measure these things; a mesh struggles to represent them.

That is the class of problem Gaussian splatting exists to solve. It describes a captured scene as a field of oriented primitives rather than a surface, which handles thin structure, transparency, and view-dependent reflection in a way triangles cannot, and renders it back from any camera position.

A point cloud as the common reference frame

On a film with many locations and many vendors, the most valuable property of scan data is that everyone is measuring from the same thing.

Layout, matchmove, environment, and lighting departments each need spatial information, and historically each derived it separately — from set drawings, from survey notes, from photographs, from memory. Those sources disagree, and the disagreements surface late.

A registered point cloud is one authoritative record. It fixes scale, orientation, and the relationships between elements once, and everything downstream inherits them. Geometry destined for rendering is converted through retopology and UV layout; geometry destined for reference stays as measured data.

Additional film scanning work is collected in the portfolio, next to the full service list.