| name | whistlegraph-pop |
| description | Download a whistlegraph TikTok and read its musicality (tempo, key, whistled-melody note sequence) as the front of the /pop track pipeline. Use when @jeffrey says "grab this whistlegraph", "analyze this tiktok's audio", "turn this whistlegraph into a pop track", or pastes a tiktok.com/@whistlegraph link. |
whistlegraph → /pop
Turns a whistled TikTok into a musical readout you can compose a /pop
track from. Tooling lives in toolchain/whistlegraph/ (see its README).
Steps
-
Grab + analyze. From a tiktok URL (or --latest N from the
@whistlegraph account):
node toolchain/whistlegraph/grab.mjs <tiktok-url>
This downloads the mp4, extracts a mono WAV, and runs analyze.py
(pop .venv librosa) → tempo, key/scale, whistled-melody note
sequence, onsets → downloads/<id>.analysis.json + INDEX.json.
-
Read the melody musically. The pyin track quantizes a portamento
whistle, so expect neighbor-note jitter (the centsOff column shows
the glides) — read the skeleton, not every blip. Identify the hook,
its repeats (titles like "… x2" mean the phrase repeats), the longest
held notes (structural anchors), and the cadence.
-
Hand off to /pop. Use key, tempoBPM, and melody[] to compose
bottom-up from AC instruments — never Suno end-to-end (see
pop/SCORE.md, pop/lib/). The whistled line is the lead/topline.
Notes
*.mp4/*.wav are gitignored (reproducible from URL); *.analysis.json
and INDEX.json are tracked.node toolchain/whistlegraph/grab.mjs --list summarizes the corpus.
