The short version: We analyzed six months of ClipMusic recognition logs — 76,017 identification attempts, 26,901 successful matches, 14,753 distinct songs. The findings: 71% of songs were identified exactly once, no song reached even 0.3% of total demand, and roughly one in five matches points at a slowed, sped-up, or remixed version instead of the original. The music people search for looks nothing like the charts.
Streaming charts measure what people play. Our logs measure something different: the moment a person hears a song in a short video, can't find its name anywhere, and cares enough to paste the link into a recognition tool. That's a stronger signal than a passive stream — and aggregated over six months, it draws a picture of short-video music that we haven't seen published anywhere else.
This post is that picture. All figures cover February 7 through August 8, 2026, from ClipMusic's internal logs.
The Headline Numbers
| Metric | Value |
|---|---|
| Identification attempts | 76,017 |
| Successful song matches | 26,901 |
| Distinct songs identified | 14,753 |
| Songs identified exactly once | 10,473 (71%) |
| Songs identified 25+ times | 41 (0.28%) |
| Share of demand held by the top 38 songs | 5.3% |
Finding 1: There Is No "Hit" — the Long Tail Is Almost Everything
The single most surprising number: 71% of all songs in our database were identified exactly once. One person, one video, one match — never seen again.
Even at the very top, concentration is astonishingly low. Our 38 most-identified songs together account for just 5.3% of all successful matches. Compare that with any streaming chart, where the top 40 tracks routinely swallow a double-digit share of total plays. Short-video sound demand behaves like the opposite of a hit economy: a vast, flat ocean of one-off curiosities, with barely a bump where the "hits" should be.
Why? Because the recognition use-case filters out everything easy. A charting song on TikTok has a labeled sound page — nobody needs a tool for it. What's left is the entire rest of the iceberg: obscure edits, regional tracks, decade-old deep cuts, game soundtracks, and bedroom-producer uploads that will never touch a chart. That iceberg turns out to be enormous.
Finding 2: Roughly One in Five Matches Is a Modified Version
We tagged every matched song whose title declares itself a variant — "slowed", "reverb", "sped up", "nightcore", "remix", "edit", "montagem" and similar markers — and weighted by identification volume:
| Version type (by title) | Identifications | Share |
|---|---|---|
| Original | 21,788 | 81.0% |
| Slowed / reverb | 3,078 | 11.4% |
| Other remix / edit / montagem | 1,848 | 6.9% |
| Sped up / nightcore | 187 | 0.7% |
That's 19% of all demand pointing at a version that isn't the original master — and this is a floor, not a ceiling, because plenty of modified uploads keep the original title. It confirms with data what we argued in our remix-culture analysis: for a meaningful slice of short-video music, the "song" people are looking for exists only as a derivative, often absent from Spotify and invisible to standard fingerprinting built against official releases.
Finding 3: The Metadata Desert
Of all successful identifications, 37% matched songs that carry no genre tag at all in the commercial metadata ecosystem. These aren't broken records — they're tracks so new, so regional, or so far outside label pipelines that nobody ever filed paperwork for them.
Where genre data exists, the demand ranking looks like this: Hip-Hop/Rap (2,639 identifications), Pop (2,510), Electronic (1,901), Dance (1,548), Alternative (1,251), then a long tail through Rock, Soundtrack, R&B and Baile Funk. Electronic plus Dance together nearly match Pop — a distribution skewed far more electronic than any consumption chart, which fits a medium where instrumental texture matters more than lyrics.
Finding 4: What the Failures Say
26,901 successes came out of 76,017 attempts. The gap is its own dataset: dead or region-locked video links, platform anti-scraping walls, clips where music sits under two layers of voice-over, and audio modified far enough that nothing matches. We break down exactly where identification fails — and how we claw some of those back — in a companion post: Why music recognition fails on short videos.
What This Means
- For listeners: if you can't find a song, it's usually not you. Odds are the track is a one-off edit or an unlabeled regional release that search engines simply don't index.
- For creators: the data rewards distinctive, obscure sounds. 71% of identified songs were looked up once — but that lookup is the highest-intent signal a track can earn, and clusters of them are how montagem edits and catalog deep cuts break out. Our monthly charts track exactly those clusters.
- For the industry: a fifth of recognition demand targets derivative versions that mostly sit outside official catalogs. That's unmonetized demand at scale.
Curious what song is behind a specific video? Paste any TikTok, Reels, Shorts, or X link — it takes about 30 seconds.
Identify a Song NowMethodology: all figures from ClipMusic internal recognition logs, Feb 7 – Aug 8, 2026. "Identification attempts" counts completed pipeline runs, including failures; "successful matches" counts user-submitted videos matched to a track. Version-type analysis is title-based and therefore conservative. Genre data joins our catalog against commercial metadata sources (Spotify / Apple Music). Nothing in this post involves personal data — counts are aggregate. Details on our About page.