Synthetic Highlights and the Rights Holder’s Dilemma
Automated clipping and generated commentary have made near-zero-cost derivative content possible at scale. The licensing framework assumes a human decided what to publish.
Photograph: Ben2742 · CC0 · Wikimedia Commons
Highlight production was until recently a constrained activity. A human editor reviewed a fixture, selected passages, assembled a package and published it within a contractually defined window at a contractually defined length. The constraint was labour, and the licensing framework was built around it: rights agreements specify permitted clip duration, the delay before publication, the platforms permitted and the territories covered.
Every one of those provisions assumes that producing a clip costs something. That assumption no longer holds.
Volume as the new variable
Current systems ingest a match feed, detect events, apply narrative rules and output finished packages — with cuts, replays, graphics and generated commentary — within seconds of the passage occurring. The marginal cost of an additional variant is effectively zero.
This makes possible a category the licensing framework has no language for: per-player packages, per-supporter-preference packages, packages assembled around a specific tactical theme, packages in a specific language with a specific commentary style, generated on demand for an individual viewer.
A rights agreement permitting a licensee to publish ninety seconds of highlights two hours after full time does not clearly address whether that licensee may publish four thousand distinct ninety-second packages, each personalised, each individually compliant with the duration limit. On the literal terms, most agreements permit it. On the evident intent of the parties when they signed, almost none do.
The contract limits the length of a clip. Nobody thought to limit how many of them there could be.
The substitution risk
The commercial danger is the erosion of the live product. Highlights have always been permitted on the basis that they are complementary — a summary that promotes the full match rather than replacing it. Duration limits and publication delays exist specifically to preserve that relationship.
A sufficiently comprehensive set of personalised clips, delivered close to real time, begins to function as a substitute. A viewer receiving every meaningful passage involving the players they care about, continuously through a fixture, has consumed something close to the match without acquiring the right to watch it.
Rights holders are responding by rewriting clip provisions around aggregate volume, total derivative minutes per fixture and minimum publication delays that scale with the volume published. These are sensible amendments and they are being negotiated into agreements one at a time, which means the position differs by competition, by territory and by licensee for the next several years.
Generated commentary is a separate problem
Synthetic commentary raises issues the clipping question does not. Systems now generate play-by-play audio in multiple languages, and several can render it in a voice closely resembling a specific commentator.
Where that voice is licensed from its owner, the arrangement is straightforward. Where it is a close approximation produced without agreement, it is not, and the legal position varies considerably between jurisdictions — some of which have specific protections for voice and likeness, and some of which do not. Broadcast unions in three markets have negotiated explicit consent requirements into recent agreements, which is the sensible route and covers only their members.
A separate question sits underneath: whether a generated commentary track constitutes a new work with its own authorship, and if so whose. Nobody has answered this and the money involved is now large enough that somebody will litigate it.
The asymmetry to watch
The most durable consequence may be distributional. Producing this content at scale requires ingestion infrastructure, trained models and rights-cleared archives. Large rights holders and large licensees have all three. Smaller competitions have none of them, and their content is precisely what is most easily appropriated by unofficial operators who face no licensing constraint at all.
The result is that the technology strengthens the parties that were already strong and increases the rate at which value leaks away from the parties that were already weak. That has been the pattern of every distribution shift in sports media for forty years, and there is no obvious reason this one will differ.