IBM Patents an AI That Cuts Personalized Highlight Clips From Live Events
Imagine a sports app that automatically sends you a clip of every moment your kid touched the ball during a school game, while sending a completely different set of clips to every other parent watching the same stream. That's the core idea behind IBM's latest patent.
What IBM's personalized event clipping system actually does
Think about the last time you watched a long recorded event and had to scrub through hours of footage to find the two minutes that actually mattered to you. IBM is patenting a system designed to do that scrubbing automatically, and differently for every person watching.
The system pulls in a recording of an event, then looks at a profile of each viewer, including details about who they are and what their connection to the event is. An AI model compares that profile against the full footage and picks out the specific moments most relevant to that particular person. You might get a five-minute clip; someone else watching the same original video gets an entirely different five minutes.
The patent covers both active participants (someone who was actually at or in the event) and passive participants (people watching from outside). The system can also push an alert to your device when your relevant segment is ready, rather than making you go looking for it.
How the ML model matches footage to each user's profile
At its core, this is a personalized video segmentation system. The computer ingests raw media content tied to an event, then separately retrieves a profile for each user. That profile is described as a "set of attributes," which likely means things like role at the event, preferences, or identity.
A machine learning model (an algorithm trained to recognize patterns) is then run on both the video and the user's attribute data together. The model identifies which portions of the recording are statistically most correlated with that user's profile, and those portions become a personalized clip.
The patent makes a specific distinction between two types of users:
- Active participants: people who were physically part of the event (athletes, speakers, performers)
- Passive participants: people who watched or attended but weren't directly involved
Once a relevant segment is identified, the system either delivers it directly to the user's device or sends an alert pointing them to it. The claim language suggests both delivery modes are covered, and that multiple users can be processed in parallel, each getting a different output from the same source footage.
What this means for sports, concerts, and enterprise meetings
The most obvious application is sports and live events, where a single broadcast contains thousands of individual moments, but each viewer only cares about a small fraction. Parents at youth sports leagues, fans following a specific athlete, or conference attendees who only care about one session would all benefit from automated, personalized clipping rather than manual editing.
There is also a clear enterprise angle. Corporate meetings, training sessions, and town halls are routinely recorded, and different employees have different stakes in what was said. A system like this could theoretically surface only the segments relevant to your team or role. IBM's long history in enterprise software makes that direction a natural fit for where this technology might actually land.
This is a real and practical idea, but it is not a novel concept in the broader market. Automated highlight generation already exists in products from Google, Microsoft, and sports-tech startups. What IBM is trying to pin down here is the specific combination of user-attribute profiling with ML-driven segmentation, applied to both active and passive participants. Whether that framing is distinct enough to matter legally is an open question, but as a product direction it is sensible rather than surprising.
The drawings
6 drawing sheets from US 2026/0196047 A1 · click any drawing to enlarge
Which company should we read for you?
We track 17 companies here. Pro is the same weekly breakdown for any company you choose, delivered privately. Type a name and we'll scope it and send you a quote.
Get one Big Tech patent every Sunday
Plain English, intelligent commentary, no hype. Free.
Editorial commentary on a publicly published patent application. Not legal advice.