日本語版 / Japanese · Detailed SVG charts · Audio prediction report · Spotify playlist predictor · Qwen re-audit
When does Yousuke Yukimatsu become shirtless?
This study reviews 23 publicly available YouTube DJ/live sets that are at least 900 seconds long. We record the first frame where Yousuke Yukimatsu is clearly visible shirtless. This is an observable confirmation time, not necessarily the physical moment when he removes his shirt.
Why—and when—does the shirt come off?
Every DJ has probably imagined, at least once, taking off a T-shirt in the spirit of Yousuke Yukimatsu. But the difficult question is not whether to take it off. It is when.
Does the shirt come off at the peak, just before it, or only after the set has crossed an invisible threshold? How is that decision related to track selection? At what moment does the heat of the room begin to demand a change in the body? These are questions many DJs may have felt, but few have ever recorded systematically.
This project examines Yousuke Yukimatsu’s past DJ and live performances, measuring when the upper body first becomes visibly shirtless and where that moment sits within the overall set. It then uses track order, track length, BPM, and energy to estimate which track—and approximately how far into it—the T-shirt would most plausibly come off in a given playlist.
This is not a manual for imitation. It is an attempt to recast an apparently personal gesture as a question of set structure, time, intensity, and track selection—and to read the relationship between physicality and music through data.
Thumbnail timeline
Each row represents one complete set. Thumbnails are placed at one-minute intervals from the start (0:00, 1:00, 2:00, and so on). The number of thumbnails follows the set length. Red outlines mark the first clearly confirmed shirtless position.
“Not observed” and “inconclusive” mean the video did not provide a reliable visual confirmation. They do not mean that he definitely remained clothed.
Method
Coarse contact sheets were used to scan the full duration. Candidate transitions were then checked with one-second contact sheets, and conflict cases were rechecked with locally decoded source frames using Qwen. The mean and median use confirmed cases only; four unobserved and six inconclusive cases are not treated as zero.
Keep Hush Live is an upper-bound first-observed time because the first checked source frame was already shirtless, so its physical transition may be earlier.
Download JSON · Download CSV · Qwen re-audit JSON · Browse all thumbnails