For Michael Kors’ Shanghai ‘The Walk’ influencer event, NYC creative agency THAT envisioned a photobooth writ large — a three-camera shoot moving down a 50’ walk against an LED wall placing the subjects in scenes shot on location around the world.
THAT wisely tapped NYC tech wizards AV&C, who knew in order to produce an edited multi-camera video, set to music with titles, every 60 seconds for 5 hours straight, they needed an inhuman editor.
For that they brought on Trash TV, a still-stealth startup developing two dovetailing machine learning softwares - one to tag video, one to edit. That provided a crucial piece of the puzzle - a near-instantaneous edit.
Beginning with Trash’s software as the cornerstone, we provided technical direction on a high-bandwidth humans-and-machines video pipeline, beginning at a user-centric check-in and stage management GUI, proceeding through an almost constantly-rolling live stage, onward to capture, ML edit, human edit pass, QA, transcode, and handoff.
With ace backend programming from Chuck Reina, we designed a stateful process management system, one where each users’ progress was managed from a handful of customized interfaces like the check-in iPads, a REST API for the automated stages, and browser-based GUIs for the postproduction team.
On the front, that user-centric flow dovetailed with a stage-focused state machine, handling video playback, video capture, and lighting, and projection on a downstage scrim to cover resets, all driven seamlessly by the stage manager’s tablet. Fortunately that dev job was dunked by the best in the biz, the irascible Torotonians, the guys who wrote the book, nVoid.
Making all that possible - the pipes - was a stack of high-availability storage with the capability to serve up 32 concurrent streams of ProRes 422 at 1080 while keeping realtime sync with a hot backup. While invisible, that combination of storage and redundancy was critical to making it possible to deliver 273 30-second videos in 5 hours. The wizard behind that curtain was AV&C’s Nate Blum.
The system was seamless and invisible, the videos were viewed by millions, and the event was nominated for a Webby and covered in WWD, Vogue, and probably a billion Chinese-language publications.
Client: Michael Kors
Production Company: AV&C
Technical Direction: Noah Norman / Hard Work Party
Machine Learning Edit Systems: Anton Marini + Hannah Donovan / Trash TV
Capture and Playback Systems: Elburz Sorkhabi + Matthew Hedlin / nVoid
Backend Programming: Chuck Reina
Engineer: Nate Blum / AV&C
Producer: Peter Stanik
Nominated: Webbys 2018
November 15, 2017