Trend Intelligence

The Trend Database: How Songbrain Planned to Track What's Viral (Offline)

May 14, 2026 · Updated October 2026 · 5 min read

Update:The trend database described below is currently offline while we rebuild it. Virality Scores today are computed from the song's own DNA — hooks, energy, structure, lyrics — with a transparent breakdown. This article stays up as documentation of how the system was designed.

The weekly loop, as designed

⏸ currently offline

Four steps, one of which nobody else does: viral_sec— not "this song is trending" but the exact second of the audio people actually clip. The loop is drawn dashed because it isn't running today.

Key takeaways

→The trend database is currently offline while it's rebuilt — Virality Scores today come from the song's own DNA (hooks, energy, structure, lyrics), not live trend data.
→The system was designed to index viral clips weekly per subgenre, down to the exact second of the audio people clip — the viral_sec field no other dataset has.
→Charts can't do this job: by the time a song charts, the viral wave is at or past its peak, two weeks too late for release timing.
→This article stays up as documentation of the design; when the database returns, trend matching plugs back into the same analysis pipeline.

As designed, every Sunday at 03:00 UTC a Songbrain worker would wake up, hit the TikTok API per subgenre, pull the top viral clips from the last week and index them. Sound-IDs. ISRC codes. The exact second of the song people clipped. The lyric snippets they pasted into captions. Hashtags. View counts. Likes-to-views ratios.

By Sunday afternoon, that would give a fresh snapshot of what's breaking in 37 subgenres — drift phonk, liquid DnB, deathcore, 4th-gen K-pop, hyperpop, indie shoegaze, modern pop, trap, drill, afrobeats, reggaeton, and more.

This was meant to be the data layer that turns Songbrain's analysis from “here's a Virality Score in a vacuum” into “here's how your song fits the trend window your subgenre is in right now.”

Why charts don't cut it

Billboard, Spotify Charts, Apple Music Top 100 — they all share the same flaw: they tell you what already won. By the time a song hits a chart, the viral wave is at or past its peak. If you're trying to time a release against the trend, the chart is two weeks too late.

Even “trending sounds” pages on TikTok and Instagram only show that a soundis trending. They don't tell you the momentin the sound that people clip. They don't break it down by subgenre. They don't expose the lyric snippets in the captions.

What the Trend Database was built to store

subgenreWhich of the 37 subgenres this clip belongs to (e.g. drift_phonk, deathcore, 4th_gen_kpop).

sound_idTikTok's internal sound-ID — lets us tie multiple viral clips back to the same audio.

isrcWhere available, the canonical ISRC of the underlying song.

viral_secThe exact second of the song where the viral clip starts. Sec-precise. This is the secret-sauce field — no other dataset has it.

lyric_snippetsLyric fragments that appear in the captions of viral clips. The line your subgenre actually hooks on.

hashtagsHashtags used by the viral clips. Frequency-weighted across the scrape.

engagementViews, likes, comments, shares, and a derived velocity score (views per day since posting).

How it was meant to feed your song analysis (not live today)

The plan: when you upload a track, cross-reference your song's hook map against the latest trend snapshot for your subgenre and answer three questions. Today's Virality Score does not do this — it is computed from the song itself.

→

Is your hook timing on the trend window?

If viral clips in drift phonk currently start between 0:10 and 0:22, and your best moment lands at 0:14, you're in the pocket. If your hook is at 1:30, you're building intros for a feed that doesn't scroll past 8 seconds.

→

Are your lyric repetitions in line with the captions people post?

If your strongest repeating line happens to match the lyric-snippet pattern in your subgenre, your captions write themselves.

→

Is your sound DNA early, on-time, or late?

Your song's CLAP embedding gets compared to the embeddings of this week's viral subgenre clips. The idea: say whether your track is leading the wave (early), riding it (on-time), or chasing it (late).

Why we built it ourselves

There's no API for this. The TikTok trends endpoint is opaque and locale-locked. Spotify's “Today's Top Hits” is a curated playlist, not raw data. Chartmetric exists, but it indexes finished charts — not sec-precise clip starts.

So we built the scrape ourselves, with a six-layer cost guard and a strict weekly cadence (Sundays only, 03:00 UTC, kill-switch enabled by default in case anything looks wrong). It is switched off while we rebuild it.

What you can use today instead

Your song analysis. Virality Score with a breakdown, Best Moments to the second, and 7 ready-to-post reels cut from them.
The Viral Radar charts. A public chart per genre of real songs analysed by Songbrain, ranked by score — browse the genres. It ranks Songbrain songs; it is not TikTok trend data.
Your own numbers. The Performance tab tracks how each reel you post did, per song.

What the Trend Database doesn't do

Even when it's back, it won't predict the next viral genre — anyone who promises that is either lying or selling a course. It was designed to track what's breaking, not what's coming.