Moodlist

Music Maketh · iOS Application · Creator

Project Summary

Moodlist is an iOS application for creating playlists from multimodal inspiration. Instead of starting only with songs, users can begin with any creative signal—songs, albums, playlists, saved moodlists, vibe text, sketches, photos, recorded audio, or Shazam-detected music—and transform that input into a coherent playlist and cover artwork.

The product is built around a structured generation pipeline rather than a simple “AI playlist” prompt. Inspiration is normalized into musical signals, expanded through search planning and retrieval, ranked through deterministic musical scoring and model-assisted reranking, and synthesized into a final playlist that users can edit, save, or export to Apple Music.

Moodlist behaves less like a one-shot generator and more like a creative workspace: drafts persist across sessions, results remain editable, and saved moodlists can be reopened or cloned into new generations.

Highlights

Multimodal creative input

Moodlist allows users to begin a playlist from multiple forms of inspiration rather than a single text prompt. The app supports songs, albums, playlists, vibe descriptions, sketches drawn with PencilKit, up to twelve visual references, short recorded audio clips, and live Shazam recognition. These inputs are normalized into structured musical signals that guide the recommendation engine.

Hybrid recommendation engine

The generation system blends deterministic music scoring with model-assisted reasoning. The pipeline extracts signals such as genre, mood, instrumentation, ambience, tempo, era, and lyrical themes, then retrieves candidate tracks from Apple Music, a seeded song catalogue, and semantic vector search. These candidates are ranked through musical feature alignment, retrieval evidence, freshness, and personalization before optional LLM reranking synthesizes the final playlist.

Background learning from the user’s library

A background seeding engine continuously analyzes the user’s library and playlists, generating reusable musical signals and search queries that expand the system’s retrieval space over time. This means the app improves its ability to find musically relevant tracks without requiring the user to restate their tastes every time they generate a playlist.

Persistent creative workspace

Moodlist treats playlist generation as reproducible creative state rather than ephemeral output. Draft sessions auto-save, users can work across multiple draft tabs, and generated playlists store immutable input snapshots that allow them to be cloned or remade later. Feedback on recommendations is persisted with full lineage data so the system can learn how tracks were chosen and which ones users preferred.

On-device AI generation

Standard playlist generation runs primarily on device using a Core ML text model, enabling fast response times and reducing reliance on external APIs. Optional Pro generation can switch to cloud models for deeper search planning and reranking when available.

Why it matters (for clients)

Moodlist demonstrates how complex creative software can combine modern AI with deterministic engineering. The system integrates multimodal inputs, hybrid recommendation logic, background learning, and responsive mobile UX into a cohesive product that feels intuitive despite the complexity underneath.

For clients building AI-powered applications, it showcases how to design systems where models augment structured pipelines rather than replace them—producing results that are explainable, controllable, and user-friendly.

‍