Publishing a trance music catalog to TikTok, Facebook and Instagram — automatically, every day

Managing a music catalog across social platforms is pure operational overhead: clip selection, formatting per platform, scheduling, hook testing, performance tracking. This stack eliminates all of it — the catalog publishes itself, the hooks evolve based on real data, and the YouTube channel visuals are generated on demand by Grok Imagine.

TikTok Studio: the central orchestration hub

TikTok Studio (Node.js/Express, port 8787, VPS) is the backbone of the publishing operation. It exposes a REST API consumed by cron jobs firing at four fixed times per day (09:00, 13:00, 18:00, 21:00 Brussels time). For each slot, it selects the next track from the queue, generates a platform-optimised clip, applies the chosen hook, and publishes via the Blotato API — a single authenticated REST call that handles TikTok, Instagram, Facebook and LinkedIn formatting, scheduling and multi-account management simultaneously. The frontend (tiktok-dashboard, Next.js 15, TailwindCSS, Recharts, tiktok.neurolinks.be) provides the hook bank editor, catalog management, episode queue and real-time analytics.

Hook selection: a bandit algorithm, not intuition

The opening three seconds determine whether a viewer stays or scrolls. Hook selection runs on an ε-greedy bandit algorithm against a bank of 80+ variants. Each variant accumulates a performance score based on real like and retention data. The algorithm exploits the best-performing hooks 90% of the time and explores new variants 10% of the time. Every week, an LLM reviews the performance data, proposes new hook variants and retires those below the threshold. Hook performance improves week over week without any manual A/B testing.

Aurora Trance & Euphoria Trance: two autonomous YouTube channels

Two trance music YouTube channels — Aurora Trance and Euphoria Trance — operate fully autonomously. For each upload: Grok Imagine (xAI) generates the cover art and background visuals from a scene prompt; OpenMontage (Python video framework) + FFmpeg assembles the track visualization; the licensed audio (795 MB local catalog) is mixed in; the final video is published to YouTube via the Data API. Both channels post on a rolling schedule with no human input. The visual style — cosmic, atmospheric, no text overlays — is enforced at prompt level by a guard function that strips any text instruction before sending to Grok.

Spotify Neurolinks: playlist automation with Claude Haiku

spotify.neurolinks.be (React/Vite, VPS) manages two Spotify channels: Music (daily at 20:00) and Tranceportation (daily at 21:00, trance-only). New releases are discovered via Scrapling, evaluated for relevance by Claude Haiku — fast, cheap, accurate for binary classification — and added to the publication queue. A deep-link (?play=track_id) triggers autoplay on the landing page, converting social traffic into Spotify plays.

Token cost vs. Belgian human equivalent

A social media manager handling catalog publishing across four platforms (scheduling, clip creation, formatting, performance tracking) costs €3,500–4,500/month gross in Belgium. A graphic designer producing channel visuals for two YouTube channels (thumbnails, backgrounds, artwork) costs €1,500–2,000/month. Total human cost: €5,000–6,500/month. The AI stack: Grok Imagine for YouTube channel artwork (60 images/month) ≈ €6/month; Claude Haiku for Spotify track classification ≈ €1/month; Blotato API for cross-platform publishing ≈ €50/month allocated; VPS compute ≈ €4/month allocated. Total: approximately €61/month. Against the €5,750/month human baseline, the AI stack reduces cost by 98.9%. The four-times-daily publishing cadence — 120 posts/month across four platforms — would require multiple people to sustain manually. Here it runs as a cron job.