From MP4 upload to processed catalog, no human in the loop.

A music distribution platform — about 5,000 songs across many independent artists — was running their upload processing across three tools: a Zapier job that copied each MP4 from a shared Google Drive folder into the right artist subfolder, an older n8n workflow that logged the file to a Google Sheet, and a manual studio step where the operator would dedup-check, run a pitch-preserving tempo stretch, and add reverb on the output. The chain worked, but it broke if any one step needed a fix, and it didn’t scale: every new artist meant more manual processing, more zaps, and more places to lose track of a duplicate or a bad upload.

We replaced the entire chain with a single self-hosted n8n workflow on a small Hetzner VPS — 58 nodes from ingest to logged catalog row. Every five minutes the pipeline scans the shared Drive root for new MP4s, runs each through ACRCloud deduplication against the artist’s own library (never the global commercial catalog, per the client’s hard requirement), checks for AI-generated audio fingerprints, applies a pitch-preserving tempo stretch with FFmpeg’s rubberband filter to land between 5:00 and 5:10, layers a SoX Freeverb reverb on the output, extracts a thumbnail at a random frame, packages a clean WAV, and uploads three artifacts back to Drive — all while writing a row to the artist’s own tab in a shared Google Sheet at every status transition. The dedup library is self-populating: every successfully processed song uploads to ACR’s custom bucket, so the next match is automatic.

The platform now processes its 5,000-song catalog through one autonomous pipeline, no manual studio step, no Zapier orchestration, and no operator dedup checks. New artists onboard by creating a Drive folder and a Sheet tab — the workflow handles the rest. The whole system runs on a $14/month VPS.