A client story
The pipeline we built around one client.
How a single enquiry became a standing delivery pipeline: media in, broadcast-ready cue sheets back.
Where it started
One of our clients came to us with episodes that were already made, moving onto a mainstream service that required cue sheets for every one of them. The programmes were finished long before anyone knew the sheets would be needed: much of the music was drawn from subscription music services, and there was no cue list that could simply be tidied up and submitted. In some episodes the music was dense, and the deadlines were real.
We said what we always say: send us what you have.
Understanding the requirements
Subscription services music raises questions that traditional library tracks do not. Who are the writers and publishers of record? Which identifiers does the recipient need to see? How are royalties handled? Before building anything, we worked through the delivery specification and the services' own catalogue data, worked out exactly how this music should be reported, and built the template their recipient expects, down to the field.
That groundwork is the part most people skip. It is also the reason everything after it became simple.
Teaching the pipeline to listen
The bigger problem was identification. With no reliable cue list, the only trustworthy source of truth was the audio itself. So we built audio-based recognition into our pipeline: fingerprinting that identifies the music straight from the finished mix, including cues sitting well under dialogue. From there the pipeline runs in four stages: reliable music recognition, then manual fingerprint verification, then metadata enrichment against the catalogues, finally sheet creation in the recipient's exact format. A person still checks every sheet before it leaves.
The brief behind the build was simple: keep their costs down. Automating the heavy lifting is what holds the per-episode cost low, and the pipeline is built to point the human straight at the weaker points of the recognition, the cues where fingerprinting is least certain, rather than re-checking what it already got right. The time is spent where it changes the outcome.
The first episodes taught us their edge cases, and every edge case became a rule in the pipeline. By the time the early deliveries were done, the unusual had become routine.
From their archive to our desk
We also took the friction out of moving the material itself. Together we set up the file handling end to end: secure large-file transfer from their source, agreed naming and folder conventions, and a drop point on our side that feeds the pipeline directly. An episode leaves their library and arrives with us ready to process, with pre-release media handled confidentially throughout.
Where it is now
Today it is a standing arrangement. An episode is chosen for delivery to the service, the media arrives, and a broadcast-ready cue sheet comes back, usually within 24 hours, in the exact format the service requires. No chasing, no reformatting, no bounced deliveries.
And the capability we built for them outlived the project: audio fingerprinting is now part of the service for every client. When no cue list survives, the audio is enough. Read more about cue sheets from the final mix.
The best measure of the pipeline is what happened to the conversation: there barely is one. Cue sheets stopped being something they think about.