What it does
Highlight reels are the currency of getting scouted, and making one is the single biggest barrier for a junior athlete. It means hours in an editing app, or paying someone, or having a parent with the time and the software. Most kids have none of those, so most kids never post, and the footage sits on a phone.
Neylo removes the entire step. Film the game on whatever is available, upload the full file, and the model does the rest. It finds the athlete, follows them through every phase of play, pulls the passages they are involved in and cuts them into a reel ready for their profile. No timeline, no trimming, no editing skill required.
For athletes
Every match you play becomes evidence instead of a memory. There is nothing to learn and nothing to buy. The athlete whose parent films on a phone gets the same quality of output as the athlete whose club owns a camera system, because the model now does the work the editing budget used to do.
For scouts and clubs
Consistency is the quiet value here. Reels on Neylo follow the same shape because the same model cuts all of them, which means comparing two athletes is comparing football rather than comparing editing. And because every clip is cut from a full match, a scout can always step back into the surrounding passage of play for context. A highlight without context is marketing. A highlight with context is scouting.
Where it fits
Neylo exists because Australia has a visibility problem in sport, not a talent problem, and the biggest source of that invisibility is raw footage nobody has time to cut. Highlight generation turns the phone on every sideline in the country into scouting infrastructure. It is also the engine that feeds the rest of the platform: the reels power profiles, and the tracking data behind them powers TalentRank.
How it’s built
The pipeline is capture-agnostic by design. Phone footage, club cameras and consumer sports cams all land in the same ingestion layer, are transcoded and stabilised into one working format, and only then reach the model. Detection finds every player in frame, keypoint tracking locks onto the target athlete and re-identifies them after occlusion, and event classification decides which passages matter.
It is the most compute-hungry thing we do. Training and inference run on GPU infrastructure on Google Cloud, and the engineering is led by our CTO Isaiah Monteleone with a five-person engineering team from the ANU’s TechLauncher program, split across dedicated workstreams covering the vision model, the processing pipeline and app integration.