Clearing Sports Audio with Ortana and ScorePlay

ScorePlay centralizes, distributes and classifies digital assets — photos, videos and live broadcasts — for more than 300 of the world's leading sports organizations. Ortana Media Group, the systems integrator behind the Cubix workflow platform, has connected two AudioShake models to it: the Copyright Compliance System for the library, and Dialogue RT for the live feed.

Sports audio is layered and inconsistent. In any given broadcast, stadium PA music, commentator dialogue and crowd noise all reach the same microphone: a celebration with an unlicensed track underneath, a sideline interview with the PA over the top, a commentator buried under the jeers after a game-winning play. That makes finished content hard to distribute, and it carries the risk of takedowns and copyright fines. Ortana addresses both problems before publication — one on assets already in the library, one on the feed as it arrives.

Music removal across the library. A social producer has a clip worth posting and an unlicensed track sitting under it. Today it goes to an editor and comes back tomorrow, past the window where it was worth the most, or it never goes out at all. With the integration the producer selects the asset in ScorePlay and gets back a version without the commercial music, commentary and crowd still in the mix. The same workflow runs against anything in the library, so a match from five seasons ago becomes redistributable to a new platform, territory or audience.

AudioShake's Copyright Compliance System detects, identifies and removes copyrighted music from a finished mix, including audio where the music sits under dialogue, crowd noise and effects. Dialogue, ambient sound and effects are preserved, and nothing is generated: what remains is recovered from the source recording. It reaches the ScorePlay pipeline through the AudioShake API, which gives full model access with no hardware requirement, at catalog scale. When an operator selects an asset, Cubix Connect and Ortana's Media Aware Workflow Engine call the API, track the job through to completion, and write the separated version back into the library alongside the original with its metadata intact.

Voice isolation on the live feed at 11ms. Live is a different problem. A broadcaster needs a clean commentary track off an ingest while the match is still running — for a syndication partner, for captioning and ASR, or for a localized feed — and cleaning the audio in post is too late for that window. In ScorePlay this has always been a VOD job. With Ortana and AudioShake it happens on the feed.

AudioShake's Dialogue RT takes the live mixed feed and returns two independent streams: an isolated dialogue stem and a separate background stem. It isolates the speech rather than suppressing what surrounds it, end to end at 11ms, and it reaches Cubix through the AudioShake SDK. Cubix ingests the feed over SDI, SRT, RTMP, NDI or HLS, writes it to the customer's own cloud storage, passes the audio to the SDK on the way through, and keeps the returned tracks co-timed with the original. Both are delivered into ScorePlay as a growing file over the platform's public API, so they appear as selectable audio in the player while the event is still running. An operator works the live timeline, selects a channel, and cuts for social or for syndication without waiting for the file to close.

Where it applies:

• Rights holders. Most of what a rights holder produces is made to go somewhere else — a broadcaster, a syndication partner, a sponsor, a league content exchange — and a clip with unlicensed music embedded either waits for clearance or never reaches the partner paying for it.
• Athlete and social distribution. ScorePlay already delivers players the media they appear in. With the music removed and the crowd noise left in place, a clip can go out the same night from the athlete's own account.
• Archive owners. Footage an organization has already paid for becomes redistributable to a new platform, territory or audience.
• Live production teams. The isolated voice track supports captioning, ASR and localization, as well as a clean commentary track for a syndication partner.

Neither platform had to build anything for the other, and the workflow is assembled through configuration rather than custom development. That is the point: audio decides whether an asset can be distributed at all, and in what window.