The Mill HoloDrive is a groundbreaking photographic location capture technique. Based on a production hardened version of Gaussian Splatting, The Mill HoloDrive allows any location to be digitally transported to any environment in post-production, providing unmatched realism and creative flexibility.
Unlike traditional geometry based 3D scanning, HoloDrive uses a photo dataset based machine learning approach to create a highly realistic, walkable photographic representation of any location. It generates a high-density, view-angle-dependent soft point cloud, offering parallax and photographic shading on captured surfaces that enhances visual realism.
With minimal on-set equipment and a low time footprint, The Mill HoloDrive captures both organic structures and hard surfaces for delivery resolutions of up to 2K* in the current version. The degree of camera movement within the HoloDrive space can be predetermined on set, allowing for versatile post-production flexibility.
Unlike server-based AI, The Mill HoloDrive handles all data processing in-house, from acquisition and model training to final pixel output. Each project is custom-trained, ensuring data privacy and production security by sourcing directly from the set itself.
Q&A
How can I imagine the output of HoloDrive?
Think of it as a “walkable photo,” where you can move the camera freely within the scene, experiencing full parallax and realistic shading. Unlike traditional photos, HoloDrive allows for camera movement in post-production, ideal for complex background shots with realistic depth and shading.
How long does a HoloDrive capture take on set?
Capture time depends on the area size where camera movement is needed in post-production (bounding box). For example, capturing a 16x10ft area with 5ft vertical freedom in our Santa Monica example took about 10 minutes.
How much processing time does a HoloDrive dataset need?
Typically, custom model training takes 2–3 days, entirely performed on our internal systems.
What scene size can HoloDrive capture?
While a dolly track-sized area is the most common capture size, HoloDrive can be extended to accommodate crane-level movement or larger scenes. Contact us to discuss aerial or expanded shots.
Does the set need to be static during capture?
HoloDrive works best on a mostly static set, similar to a 10-minute long exposure. We’ve optimized the process for stability, but minimal movement is ideal.
Can HoloDrive capture motion?
No, HoloDrive captures a static, walkable scene that allows for camera movement in post, but it doesn’t capture motion. However, moving elements can be added in post-production through The Mill’s award winning VFX pipeline.
Can lighting be adjusted in HoloDrive scenes during post?
No, lighting is “baked in” during capture, similar to traditional photography.
Does HoloDrive require a blue or green screen setup?
No, HoloDrive captures the environment as it is. However, blue or green screens may be needed for foreground elements.
Can I alter elements in a HoloDrive scene?
Yes, elements can be repositioned within the scene, similar to stamping in Photoshop, though they can’t be rotated. Scenes can also be further enhanced using The Mill’s award-winning VFX pipeline.
How should I shoot foreground elements?
Foreground elements can be shot before or after the HoloDrive capture with any lens and camera setup. HoloDrive adapts to these in post-production, but lighting should match the HoloDrive scene to achieve seamless compositing.
Is HoloDrive secure for data-sensitive projects?
Yes, all data acquisition and processing occur offline on our physical hardware, ensuring data security.
How quickly can HoloDrive be deployed?
With a minimal on-set footprint, HoloDrive requires only one operator and a single Pelicase, making it highly portable and suitable for air travel.