THE ENGINEERING OF RADIO GRANELLÌK: PROPRIETARY AI ENGINES THEMA & LYRA
How over 1,300 unreleased original tracks, 10 humanized AI speakers, and an uninterrupted 24/7 audio-visual broadcast are created: the proprietary technology that makes Radio Granellìk a world-first.
Almost all so-called "AI radio stations" merely stream automated commercial playlists or employ generic chatbots to read weather reports. Radio Granellìk is fundamentally different. Every single track, station ID, soundtrack, or videoclip on air is an original piece conceived, arranged, and written by two bespoke software systems engineered by Stefano Pastori: THEMA (the acoustic and cultural sound design architect) and LYRA (the authored vocal, metric, and cinematic directing simulator).
🏛️ THEMA — Sound Design Architect & VCSD AUDIO & CINEMATIC
THEMA (Video Context Sound Designer) is the world's first software featuring Cultural Intelligence applied to musical production for video, commercials, trailers, and broadcasting. Rather than slapping superficial ethnic samples over generic backings, it completely re-engineers the soundscape, harmonic progressions, and mixing to match specific target cultures and narrative contexts.
📸 THEMA Operational Screenshots (Click to expand)
🌍 True Cultural Intelligence
Covers 25+ Countries/Regions with 4 dedicated variants each. An action ad in China produces energetic C-Pop; in Japan it deploys J-Rock/Anime battle themes; in Italy it activates the melodic Sanremo orchestral style.
📜 92+ Arrangement Recipes
Every subcategory (`thema_arrangement_recipes.py`) features strict configurations for Soloists (Lead 1 & 2), Foundation (dedicated Bass and Drums), Production DNA, and negative filters (avoid) to prevent genre bleeding.
⏱️ Silent CHRONOS Notation
Mathematical injection of silent punctuation strings (. , !) at structural section tails to control BPM, breathing, and metric pacing without polluting lyrics with vocal artifacts.
🛡️ Anti Wall-of-Sound Algorithm
Strict hierarchical separation between primary/secondary soloists and background rhythmic foundation, preventing frequency collision and acoustic masking.
✨ Prisma Creativity Matrix
Dynamic 3-tier matrix (Safe, Extended, Exotic) with a curated Spice Pool to modulate creative variance while maintaining total stylistic integrity.
👁️ AI Multimodal Ingest
Direct analysis with Gemini from images, audio tracks, video files (via ffmpeg), and video URLs (via yt-dlp) to extract emotional temperature, energy, and BPM.
🎙️ LYRA — VJM Simulator & Lyrics Engine VOCAL & AUTHOR
LYRA (Visual Jukebox Manager) is the authored lyrical writing, vocal sculpting, and cinematic directing simulator. Designed to transcend repetitive commercial AI lyrics, LYRA enforces a strict authoring protocol with metacognitive control over voice, meter, and structure.
📸 LYRA Operational Screenshots (Click to expand)
✍️ VJM Kernel Protocol (Authored Writing)
Strict writing constraints: tech/corporate jargon decontamination (VDL), banned lazy rhymes like heart/start (V-ARP), and structured semantic arcs (VSA: Narrative, Impressionist, Dialectic).
🧬 Sonic DNA Database (929+ Artists)
Proprietary database cataloging 929 artists across 6 layers: clean vocal architecture, Lead 1 soloist, Lead 2 harmony, Foundation (specific Bass and Drums), Signature hallmarks, and studio console mastering.
🎭 Lyric Depth Modules (Depth 1-2-3)
Depth-1: Tangible physical image paired with inner psychological emotion. Depth-2: Dramatic suspensions (...) every 5 lines. Depth-3: Semantic callback (initial object reintroduced and transformed in the finale).
🎙️ VOX LAB & Italian Regional Dialects
Vocal anchor sculpting: Archetype, Environment, Texture, Vibrato, and verified Italian dialects with dual output (phonetic vocal delivery for the AI and correct written lyrics).
🔇 Anti-Screaming & Anti-Humming Matrix
Vocal stability engineering: eliminates clipping/screaming and whispering/ASMR triggers, shifting dynamic drive to the musical arrangement to ensure crystal-clear vocal projection.
🎬 Videoclip Director & Anima Engine
Audio silence analysis with ffmpeg, 7.0s clip storyboards for generative video AI, and the Anima Engine (26 personality profiles rooted in Elegant Cynicism) giving voice to Radio Granellìk's 10 AI speakers.
🔄 THE RADIO GRANELLÌK BROADCAST PIPELINE
How THEMA and LYRA interconnect to produce an uninterrupted 24/7 schedule of original music, spoken shows, and visual broadcasts.
1. THEMA
Cultural sound design, instrumental arrangement, BPM, and CHRONOS metric pacing.
2. LYRA
Authored poetic lyrics via VJM Kernel, VOX LAB vocal anchor, and 6-layer Sonic DNA.
3. VIDEO & SOCIAL
7.0s scene storyboard directing, cover art visual synthesis, and 9:16 vertical video export.
4. ON AIR 24/7
Broadcasting on Radio Granellìk with 10 humanized AI speakers and zero commercial ads.
EXPERIENCE THE TECHNOLOGY LIVE
Every single track playing right now on our live stream was produced using this bespoke software ecosystem.
▶ LISTEN TO THE LIVE STREAM NOW