Volumetric Performance Capture and Playback: A Workflow with Python
summary
Python glues a physical rig of eight Android phones to a processing pipeline that actually produces playable volumetric shorts: a Flask web API coordinates capture and uploads, node apps on phones invoke Android phones + Open Camera, videos get frame-aligned and converted to meshes with Metashape, ffmpeg, Blender, then played back in a 3D engine (gato). A four‑second clip becomes ~1 GB/s uncompressed (~150 MB/s compressed) and processing hits about an hour per second on an older tower, but the pipeline reliably generates glitchy organic volumetric video.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:18About Project 39
- 2:04What is volumetric capture
- 2:47Agenda
- 3:24Building a rig
- 4:22Controlling the rig
- 7:46Android apps
- 10:10Taking photos
- 11:12Taking videos
- 12:24Converting to 3D
- 15:09Camera relationships
- 16:09Texturing
- 17:05Playback
- 20:26Questions
related talks
46mVolumetric Capture and Distribution in the 5G World
27mA/V Streaming Workflow in Python
25mFirefly: Untethered Multi-user VR for Commodity Mobile Devices
42mVideo Production Strategies Using Image Sequences
31mVideo recording conferences and user groups using Python
33mCinematic Virtual Reality with 360 degree movies
51mLet's Play! Build your first Cardboard App (en)
21mLeveraging Python for Real-Time Image processing