Command Palette

Search for a command to run...

UnylyUnyly
Весь каталог

Music Attribution Scaffold

БесплатноНе проверен

Research scaffold for multi-source music attribution with transparent confidence scoring. Companion code to Teikari, Petteri. 2026. “Governing Generative Music:

GitHubEmbed

Описание

Research scaffold for multi-source music attribution with transparent confidence scoring. Companion code to Teikari, Petteri. 2026. “Governing Generative Music: Attribution Limits, Platform Incentives, and the Future of Creator Income.” SSRN Scholarly Paper No. 6109087. SSRN, https://dx.doi.org/10.2139/ssrn.6109087

README

Watch the 1-minute MVP demo on YouTube — confidence scoring, provenance timelines, and MCP permission queries in action.

Watch the 1-minute demo · 📄 Read the paper (supplementary material includes an annotated screenshot walkthrough)

CI Tests: 1311 Python 3.13 License: MIT DOI Code style: ruff Type checked: mypy uv Docs

Research-grade music attribution infrastructure with calibrated confidence scoring, multi-source entity resolution, and machine-readable consent.

Music metadata is broken — over 40% of records are incorrect or incomplete across databases that don't talk to each other. When generative AI adds millions of new tracks per month, knowing who made what becomes both a legal and economic emergency. This scaffold provides the open-source infrastructure to address it.

Overview of the Music Attribution Scaffold: editorial frontend with confidence gauges, assurance badges, review queue, and agentic sidebar for natural-language attribution queries across multi-source music metadata.

The scaffold unifies fragmented music metadata into confidence-scored attribution records with transparent provenance.


What It Does

Capability What It Means Deep Dive
Multi-source ETL Fetches and normalizes metadata from MusicBrainz, Discogs, AcoustID, audio file tags, and artist self-reports src/music_attribution/etl/ · Docs
Entity resolution 5-strategy cascade: identifiers → fuzzy strings → embeddings → graph → LLM disambiguation src/music_attribution/resolution/ · Docs
Calibrated confidence Per-field scores via conformal prediction — "90% confident" actually means ≥90% coverage src/music_attribution/attribution/ · Docs
Permission patchbay MCP server for machine-readable AI training consent (ALLOW / DENY / ASK) src/music_attribution/mcp/ · Docs
Agentic UI PydanticAI agent + CopilotKit sidebar for natural-language attribution queries src/music_attribution/chat/ · Docs
Voice agent Pipecat voice pipeline with 5-dimension persona, drift detection, and domain tools src/music_attribution/voice/ · Guide
Editorial frontend Next.js 15 with confidence gauges, assurance badges, review queue frontend/ · Docs
How does the 5-strategy entity resolution cascade work?

Entity resolution answers: "Is the 'Imogen Heap' in MusicBrainz the same person as 'I. Heap' in a Discogs release?" The cascade progresses from cheap/certain to expensive/uncertain:

  1. Identifier matching — Exact ISRC, ISWC, ISNI, or MBID match. Cheapest and most reliable.
  2. String similarity — Jaro-Winkler and token-sort ratio on canonical names and aliases. Handles typos and transliterations.
  3. Embedding similarity — Sentence-transformer embeddings compared via cosine similarity. Catches semantic equivalences that string metrics miss (e.g., "紫雨林" ↔ "Jaurim").
  4. Graph evidence — Shared collaborators, labels, or recordings boost match probability via Jaccard coefficient.
  5. LLM disambiguation — For remaining ambiguous cases, an LLM evaluates contextual evidence. Cost-gated and cached.

Each stage produces a confidence score. The cascade short-circuits: if Stage 1 finds a high-confidence match, Stages 2–5 are skipped.

References: Fellegi & Sunter (1969) — foundational probabilistic record linkage model; Papadakis et al. (2021) — modern entity resolution survey.


Architecture

Five sequential pipelines connected by typed Pydantic boundary objects:

Architecture diagram: five-pipeline data flow for music attribution showing ETL ingestion from MusicBrainz and Discogs, entity resolution with fuzzy matching, transparent confidence scoring engine, FastAPI REST and MCP permission server, and PydanticAI chat interface.

  I. ETL          II. ENTITY         III. ATTRIBUTION    IV. API/MCP      V. CHAT
  PIPELINE         RESOLUTION         ENGINE
  ┌─────────┐     ┌─────────────┐    ┌────────────┐     ┌──────────┐    ┌──────┐
  │ Discogs │     │ Fuzzy match │    │ Confidence │     │ FastAPI  │    │ AI   │
  │ MBrainz │────▶│ + dedupe    │───▶│ scoring +  │────▶│ REST +   │───▶│ Chat │
  │ AcoustID│     │ + graph     │    │ provenance │     │ MCP      │    │      │
  └─────────┘     └─────────────┘    └────────────┘     └──────────┘    └──────┘
       │                │                  │                  │              │
  NormalizedRecord  ResolvedEntity   AttributionRecord    JSON + MCP     AG-UI SSE

  Cross-cutting: BatchEnvelope · DriftDetector · PipelineFeedback
  Storage:       PostgreSQL + pgvector

Each boundary crossing carries validation, confidence scores, and provenance metadata. Reverse feedback flows (dispute signals, recalibration requests) travel upstream via PipelineFeedback objects.

Full architecture docs: Architecture Overview · Backend Guide · docs/architecture/


Quick Start

Prerequisites

  • Python 3.13+ and uv
  • Docker and Docker Compose (for PostgreSQL + pgvector)
  • Node.js 18+ and npm (for the frontend)

Setup

git clone https://github.com/petteriTeikari/music-attribution-scaffold.git
cd music-attribution-scaffold

# Install deps, start Docker, run migrations, seed Imogen Heap data
make setup

# Start the backend API + PostgreSQL
make dev
# → http://localhost:8000 (Swagger at /docs)

# In a second terminal: start the frontend
make dev-frontend
# → http://localhost:3000

# Or: full agentic UI (backend + CopilotKit sidebar)
make agent
Environment variables
Variable Required Default Purpose
DATABASE_URL Yes PostgreSQL connection string
ANTHROPIC_API_KEY For agent Anthropic API key for the PydanticAI agent
ATTRIBUTION_AGENT_MODEL No anthropic:claude-haiku-4-5 PydanticAI model string
CORS_ORIGINS No http://localhost:3000 Comma-separated CORS origins
NEXT_PUBLIC_API_URL No http://localhost:8000 Backend URL for the frontend
LOG_LEVEL No INFO Python logging level
ENVIRONMENT No development Runtime environment

Create a .env file in the project root, or export directly. See src/music_attribution/config.py for all options.

Make commands reference
Command What It Does
make setup Full setup: deps + Docker + migrations + seed data
make dev Start backend API + PostgreSQL
make dev-frontend Start frontend dev server
make agent Start full agentic UI (backend + CopilotKit)
make test-local Run 687 unit + 43 integration tests locally
make test Run 744 tests in Docker (CI-parity)
make test-frontend Run 560 Vitest tests + WCAG checks
make test-all Full CI: lint + typecheck + backend + frontend
make lint-local ruff check + ruff format --check
make typecheck mypy (strict)
make test-cov pytest with coverage report
make format Auto-format with ruff

Assurance Levels (A0–A3)

Every attribution carries an assurance level indicating verification depth. These are mapped to international music industry identifiers:

Assurance badge system: four tiers from A0 (no data, unverified) through A1 (single source, ISRC), A2 (multi-source agreement, ISWC), to A3 (artist-verified identity, ISNI) showing progressive verification depth for music attribution.

Level Verification Evidence Typical Identifier Example
A0 None Self-declared "I wrote this song" (no external confirmation)
A1 Single source One database confirms ISRC MusicBrainz lists the recording
A2 Multi-source Independent sources agree ISWC MusicBrainz + Discogs + file metadata agree
A3 Identity-verified Full provenance chain ISNI / IPI Artist self-report + ISNI + multiple databases
How are assurance levels computed?

Assurance levels are determined by the entity resolution cascade and the number of independent sources that agree:

  • A0: No external data available. The system has only what the user claimed.
  • A1: Exactly one external source confirms (e.g., MusicBrainz has a matching ISRC). Confidence is moderate but un-corroborated.
  • A2: Two or more independent sources produce consistent information. Cross-source agreement significantly increases confidence.
  • A3: Identity-verified through a chain that includes artist self-report AND an identity standard (ISNI or IPI) AND at least one database match. This is the highest tier — reserved for attributions where the creator's identity is cryptographically or administratively confirmed.

The mapping between assurance levels and identifiers is project-defined and reflects industry best practices, not a formal standard.

Concept page: Assurance Levels


Confidence Scoring

Every attribution field carries a calibrated confidence score. Unlike typical ML "confidence" (often miscalibrated), the scaffold uses conformal prediction with Adaptive Prediction Sets (APS) to provide distribution-free coverage guarantees.

Conformal calibration pipeline: comparing traditional miscalibrated ML confidence with conformal prediction providing mathematically guaranteed coverage — a stated 90% confidence interval genuinely contains the true value at least 90% of the time.

What is conformal prediction and why does it matter?

Traditional ML "confidence scores" are often miscalibrated — a model might say "95% confident" when it's actually correct only 70% of the time. In attribution, this is dangerous: overstating confidence in who created a song can have legal and financial consequences.

Conformal prediction provides distribution-free coverage guarantees:

  1. Learns a nonconformity score from calibration data (how "unusual" each prediction is).
  2. At inference, constructs a prediction set that contains the true value with probability ≥ (1 - α).
  3. If α = 0.10, the resulting 90% prediction set genuinely covers the truth ≥ 90% of the time — no distributional assumptions required.

The scaffold calibrates every attribution field: confidence in creator identity, confidence in role assignment, confidence in recording-to-work linkage.

References: Angelopoulos & Bates (2025) — conformal prediction tutorial; Quach et al. (2024) — conformal prediction applied to language model uncertainty.

Concept page: Conformal Prediction


MCP Consent Infrastructure

The scaffold includes a Model Context Protocol (MCP) server that provides machine-readable permission queries for AI training rights. AI platforms can ask "may I use this recording for training?" and receive structured responses.

MCP consent infrastructure: AI platforms send machine-readable permission queries to the MCP server, receiving structured ALLOW, DENY, or ASK responses with conditions — enabling automated, auditable consent management for AI training on music data.

AI Platform                    MCP Server                    Rights DB
    │                              │                            │
    │── check_permission ─────────▶│                            │
    │   (recording_id, use_type)   │── query rights ──────────▶│
    │                              │◀── permission + conditions─│
    │◀── ALLOW / DENY / ASK ──────│                            │
    │    + conditions, expiry      │                            │

Responses include: permission status, conditions (attribution required, non-commercial only, etc.), expiry dates, and rights holder contact for ASK responses.

Concept page: MCP Consent · Implementation: src/music_attribution/mcp/


Voice Agent (Experimental)

The scaffold includes an open-source voice agent built on Pipecat (BSD-2-Clause, 10.4k stars). The same 4 domain tools available in the text agent are exposed as voice-callable functions -- ask about confidence scores, search attributions, suggest corrections, and submit feedback by speaking.

Architecture diagram of a five-stage real-time voice agent pipeline showing Transport, STT, LLM, TTS, and output stages with semantic endpointing as the orchestration layer for music attribution voice interactions.

Five-panel architecture: Transport → STT → LLM → TTS → Transport, with semantic endpointing as the orchestration layer.

# Quick start — zero API keys, fully local
uv run python scripts/voice_demo.py

# With custom providers
uv run python scripts/voice_demo.py --stt deepgram --tts elevenlabs

# With drift monitoring
uv run python scripts/voice_demo.py --drift-monitoring --verbose

The pipeline in three lines:

Mic → [Silero VAD → STT → ContextAggregator → LLM (4 tools) → DriftMonitor → TTS] → Speaker
       └─ Pipecat Pipeline ──────────────────────────────────────────────────┘

This is a scaffold, not a production voice product. Every component is swappable with a one-line config change:

Layer Default (Zero-Cost) Commercial Alternative Config
STT Whisper (MIT) Deepgram Nova-3, AssemblyAI VOICE_STT_PROVIDER
TTS Piper (GPL) / Kokoro (Apache 2.0) ElevenLabs, Cartesia VOICE_TTS_PROVIDER
Transport WebSocket SmallWebRTC, Daily WebRTC VOICE_TRANSPORT
LLM Any OpenAI-compatible (Ollama, vLLM) Anthropic, OpenAI VOICE_LLM_MODEL
Persona 5-dimension prompt-layered Letta (MemGPT), Mem0 VOICE_PERSONA_ENABLED
Drift EWMA cosine similarity -- VOICE_DRIFT_MONITORING
Guardrails Regex fallback NeMo Guardrails (Colang 2.0) VOICE_GUARDRAILS_ENABLED

What Makes This Different

  • Scaffold, not product: Branching paths for every component -- teams with different constraints (budget, latency, licensing) can instantiate different stacks from the same blueprint
  • Zero-API-key local dev: Default stack (Whisper + Piper + WebSocket) runs entirely on your machine at $0.00/min
  • Protocol-based swapping: STTServiceProtocol, TTSServiceProtocol, and DriftDetectorProtocol use Python structural typing -- implement the methods and it just works, no inheritance required
  • Persona drift prevention: 5-dimension persona architecture with EWMA-smoothed drift detection prevents the 8-turn persona drift cliff documented in the literature (Li et al., 2024)
  • Conditional imports: Config, persona, drift, and tool schemas work without Pipecat installed -- only build_pipecat_pipeline() requires the actual library. All 161 voice tests pass in both modes

Benchmarks (RTX 2070 Super)

STT inference is the latency bottleneck. GPU acceleration provides 4–16x speedup over CPU, with larger models benefiting disproportionately:

Component Model CPU GPU (CUDA) Speedup
STT (faster-whisper) tiny 801 ms 188 ms 4.3x
STT (faster-whisper) small 4,316 ms 267 ms 16.2x
Drift detection all-MiniLM-L6-v2 4.4 ms

Measured on NVIDIA GeForce RTX 2070 Super (7.6 GB VRAM), PyTorch 2.10.0+cu128, 10-second synthetic 440 Hz sine wave (16 kHz, 16-bit PCM). CPU uses int8 quantization; GPU uses float16. Model load times: tiny CPU 5.9s vs CUDA 0.4s; small CPU 41.1s vs CUDA 0.4s.

Run the benchmark yourself
# CPU + GPU benchmark (tiny and small models)
uv run python scripts/benchmark_voice.py --models tiny,small --output results.json

# CPU-only (no CUDA required)
uv run python scripts/benchmark_voice.py --cpu-only --models tiny

# With synthetic command accuracy (WER + domain keyword survival)
uv run python scripts/benchmark_voice.py --models tiny

# End-to-end pipeline with mock LLM (STT → LLM → TTS latency breakdown)
uv run python scripts/benchmark_voice.py --models tiny --mock-llm --llm-delay-ms 200

The benchmark script auto-detects hardware (CPU, GPU, VRAM), applies VRAM guards (skips models too large for available memory), and outputs structured JSON + a summary table. See scripts/benchmark_voice.py.

Step-by-step anatomy of a single voice turn traversing seven processing stages with latency budget annotations, targeting under 500 milliseconds total for natural conversational music attribution interactions.

A single voice turn: 7 steps from utterance to response, targeting <500ms total.

Full guide: Voice Agent Implementation · Alternatives table: Component Alternatives · Research base: Voice Agent Research · Figure gallery: 48 voice figures · Implementation: src/music_attribution/voice/


Sample Data

The scaffold ships with 8 Imogen Heap works as seed data, spanning the full confidence spectrum from 0.00 to 0.95:

Work Confidence Assurance Sources Status
Hide and Seek 0.95 A3 MB, Discogs, AcoustID, Artist Verified
Tiny Human 0.91 A3 MB, Discogs, Artist Verified
The Moment I Said It 0.82 A2 MB, Discogs, File Stable
Goodnight and Go 0.72 A2 MB, Discogs Stable
Headlock 0.58 A1 MB, Discogs Needs review
Just for Now 0.35 A1 MB only Needs review
2-1 0.28 A1 File metadata only Needs review
Blanket (unreleased) 0.00 A0 None Needs review

Each record carries a full provenance chain showing how confidence evolved through fetch → resolve → score → review events.

Why Imogen Heap?

Imogen Heap is both a critically acclaimed artist and a pioneer in music attribution technology. Her Mycelia project (2015) explored blockchain-based music attribution. Her discography spans well-attributed major-label releases (high confidence) through unreleased demos (zero confidence), providing natural coverage across all assurance levels.


Project Structure

music-attribution-scaffold/
├── src/music_attribution/          # Python package
│   ├── schemas/                    # Pydantic boundary objects
│   ├── etl/                        # Data source connectors + quality gate
│   ├── resolution/                 # 5-strategy entity resolution cascade
│   ├── attribution/                # Aggregation, conformal scoring, priority queue
│   ├── api/                        # FastAPI routes: attribution, health, permissions
│   ├── chat/                       # PydanticAI agent + AG-UI endpoint
│   ├── mcp/                        # MCP permission patchbay server
│   ├── voice/                      # Pipecat voice agent pipeline + persona
│   ├── db/                         # SQLAlchemy models, 8-table schema
│   ├── search/                     # Hybrid search (text + vector + graph)
│   ├── pipeline/                   # DAG runner for orchestration
│   ├── observability/              # Prometheus metrics
│   └── seed/                       # Imogen Heap mock data
├── tests/                          # 744 tests: 687 unit + 43 integration + 14 eval
├── frontend/                       # Next.js 15, TypeScript strict, Tailwind v4
├── docker/                         # Dockerfiles: dev, test, prod
├── alembic/                        # Database migrations
├── docs/                           # MkDocs site + PRD + knowledge base
└── pyproject.toml                  # Single source of truth for all deps

Each module has its own README with architecture details:

Module README What's Inside
Python package src/music_attribution/ Package overview, module map
ETL pipeline src/music_attribution/etl/ Connectors, quality gate, normalization
Entity resolution src/music_attribution/resolution/ 5-strategy cascade, Splink, embeddings
Attribution engine src/music_attribution/attribution/ Aggregation, conformal scoring, review queue
API src/music_attribution/api/ FastAPI routes, middleware, OpenAPI
Chat agent src/music_attribution/chat/ PydanticAI tools, AG-UI streaming
MCP server src/music_attribution/mcp/ Permission queries, consent model
Schemas src/music_attribution/schemas/ All Pydantic boundary objects
Frontend frontend/ Next.js 15, design system, components
Tests tests/ Testing strategy, fixtures, CI
Docs docs/ MkDocs site, figures, knowledge base
Architecture docs/architecture/ System diagrams, ADRs, security
PRD docs/prd/ Probabilistic decision network

Testing

1,304 tests across backend and frontend:

make test-local          # 744 backend tests locally (~30s)
make test                # 744 backend tests in Docker (CI-parity, ~90s)
make test-frontend       # 560 Vitest tests + WCAG accessibility checks (~15s)
make test-all            # Full CI: lint + typecheck + all tests
Test architecture
Tier Count Speed Infrastructure What It Tests
Unit 687 ~30s None (all mocked) Business logic, schema validation, algorithm correctness
Integration 43 ~60s Docker (PostgreSQL + pgvector) Database round-trips, API endpoints, search
Eval 14 ~5s None Benchmark validation, output schema checks
Frontend 560 ~15s jsdom (Vitest) Components, hooks, stores, WCAG accessibility
Smoke ~5s Docker Container health, migration success

All tests run in CI via GitHub Actions with path-based filtering — backend changes don't trigger frontend CI and vice versa. See tests/README.md.


Documentation

Resource Description
GitHub Pages site Getting started, concepts, API reference, tutorials
Concepts Oracle Problem, A0–A3 assurance, conformal prediction, MCP consent, entity resolution
API Reference Auto-generated from numpy-style docstrings (mkdocstrings)
Tutorials Reproducing the paper, adding data sources, API examples
Troubleshooting 15 common issues: Docker, database, testing, frontend, CI
PRD Decision Network 85 probabilistic decision nodes, 5 levels, 4 team archetypes
FinOps Landscape Deployment cost analysis: 5 paths, DevOps tax, cloud credits, egress optimization
Geopolitical Risk Assessment Cloud sovereignty: CLOUD Act, EU providers, Schrems III, Section 301
Knowledge Base RAG-optimized markdown: attribution papers, agentic systems research
Pitch Deck 12 investor slides: market, architecture, moat, revenue, roadmap
Advanced Due Diligence 18 deep-dive slides: technical, competitive, financial, team

Preview docs locally:

uv sync --group docs && uv run --group docs mkdocs serve
# → http://localhost:8000

Music AI Attribution Landscape

The music AI attribution space is undergoing rapid consolidation. This section provides a compact overview — see the full landscape report for detailed company profiles, academic papers, and buy-vs-build analysis.

Attribution Infrastructure

Company Approach Funding Key Differentiator
Sureel AI Training-time gradient tracking Undisclosed 86–90% accuracy claimed; STIM partnership; 5 patents
Musical AI Training-time influence tracking $6M Fairly Trained certified; Beatoven.ai integration
Vermillio Content monitoring (TraceID) $16M Sony Music's first AI investment; TIME100
ProRata.ai Cross-media attribution $40M UMG partnership; beyond music
Pex Content identification Established Positioning for AI attribution
Auracles Artist-controlled digital identity & permissions Undisclosed Founded by Imogen Heap; artist-verified metadata layer
Klay Vision Large Music Model with per-use attribution Licensed by all 3 majors First to secure UMG + Sony + WMG licensing (Nov 2025)
SoundPatrol Neural fingerprinting (semantic similarity) Undisclosed UMG & Sony collaboration; catches AI derivatives beyond exact matching
Orfium (AIXPERT) Explainable AI for copyright attribution EUR 7.5M EU grant 3-year research consortium; 3 of 135 proposals selected
Audible Magic Automated content recognition Established 25+ years; Udio partnership; "Broad Spectrum" altered-audio detection
BMAT Audio fingerprinting at scale Established 80M IDs/day; 100+ CMOs; 2,000+ labels in 85 countries
Muso.AI Verified music credits platform Established 50M+ verified tracks; identity verification via ID scan
This scaffold A0–A3 assurance + conformal prediction Open-source Transparent, auditable, MCP-native

AI Music Generation (with attribution status)

Platform Funding Attribution Status
Suno $375M+ ($2.45B val.) Settling lawsuits; licensed-only models in 2026
Udio Settled Pivoting to walled-garden fan remixing
Soundverse Undisclosed Published ethical framework; DNA Models for artist-owned AI
Boomy Established Fairly Trained certified; copyright-safe models
Stability AI Established Licensed data from AudioSparx
Jen Backed by Futureverse Fairly Trained certified; outputs vetted against 150M tracks
LifeScore Warner Music investment Provenance-tracked adaptive music; co-founded with Siri co-inventor
Rightsify Established Owns 12M+ songs for training; Fairly Trained certified; Hydra II model

Licensing & Certification

Entity Role
Fairly Trained Certification body — 19 certified entities
LANDR Fair Trade AI Pro-rata revenue sharing for opted-in artists
Lemonaide Cohort-based attribution; Fairly Trained certified
SoundExchange AI Registry ISRC-based opt-in/opt-out for AI training
STIM World-first collective AI music licence (2025)
Kits AI Ethical voice cloning with dynamic revenue-sharing
VEVA Collect Attribution-by-design — DAW plugin captures credits at recording time; Library of Congress partner
Digimarc Audio watermarking (NASDAQ: DMRC); survives compression/remixing; C2PA-aligned

AI Music Detection

Company Approach Key Differentiator
Ircam Amplify AI-generated music detector (98.5% accuracy) Model-specific: identifies which AI (Suno, Udio, etc.) produced a track
MatchTune / CoverNet Cover + AI clone detection across platforms Catches slowed, pitch-shifted, filtered versions; real-time alerts
Deezer Platform-integrated AI detection; two patents filed First streamer to label AI music; detected 60K AI tracks/day (Jan 2026); licensing tech to industry
Resemble AI Detect-2B model for AI audio (94–98% accuracy) Combines generation + detection + watermarking; watermark survives model training
Sensity AI Multi-modal deepfake detection (audio + video + image) Forensic approach; serves defense & law enforcement on four continents
Kliga Consumer-facing AI music detector (99.9% claimed) Client-side processing; accepts Spotify URLs; Show HN project

Music Metadata & Credits

Company Approach Key Differentiator
Musixmatch Lyrics + metadata database (12M+ songs, 250+ languages) Apple Music, Spotify, Google partnerships; AI licensing deals with all 3 major publishers
Cyanite AI-powered music tagging & similarity search 100+ music companies; taxonomy translation across different data standards
Bridge.audio Collaboration + metadata + sync licensing $3.28M seed (2024); captures rights splits at creation; Paris-based
OpenPlay Enterprise catalog management & rights delivery Spun out of Concord (2013); trusted by major labels; deep API platform
rightsHUB Cloud-based rights & metadata management Rights-neutral (takes no ownership); DataDoktor compliance checker

Audio Recognition & Fingerprinting APIs

Company Approach Key Differentiator
ACRCloud Cloud fingerprinting & recognition API Broadcast monitoring; content protection; API-first integration
AudD Music recognition API (neural fingerprinting) Enterprise endpoint processes hours-long mixes; Warner, UMG, Sony as customers

Voice Identity & Ethical Voice AI

Company Approach Key Differentiator
Vocalist.ai AI vocal transformation with royalty-sharing Artists share subscription revenue when their voice is used; founded by songwriter Andrew Frampton
Respeecher Consent-based voice cloning Hollywood-proven (Lucasfilm/Mandalorian); explicit consent per project; music-specific tools

Audio Processing & Stem Separation

Company Approach Key Differentiator
AudioShake AI stem separation for any recording $24.1M raised; all 3 majors; preparing audio for AI training; 100M+ minutes processed
Music.AI Orchestrator platform (50+ AI audio modules) 18-stem separation; trained exclusively on owned/licensed music; batch API

Royalty Infrastructure

Company Approach Key Differentiator
Beatdapp Streaming fraud detection $17M raised; UMG partnership; <0.001% false positive; claims fraud costs $2B/yr
Revelator Blockchain-based royalty distribution (Base, Polygon) Smart contract splits; daily USDC settlement cycle; Artist Wallet
Utopia Music Data-tracking music ecosystem Acquired Musimap (emotional AI) + Lyric Financial; global consumption data
Songtradr B2B music tech platform $30M+ Series C; acquired Bandcamp + Musicube (50M+ titles); end-to-end supply chain

Provenance & Opt-Out Infrastructure

Company Approach Key Differentiator
Spawning AI Creator opt-out from AI training datasets 78M artworks opted out; ArtStation & Shutterstock partnerships; API for compliance
Numbers Protocol Blockchain provenance for digital media Capture → Certify → Check framework; Numbers ID (Nid) for any media file
Yakoa Digital rights protection & originality detection $4.8M raised; AI maps creative evolution across media; NFT forgery detection

Standards Bodies & Research

Entity Role
DDEX Industry standard for music metadata exchange (ERN 4.3.x); AI Ad Hoc Group developing AI-specific standards
Fraunhofer IDMT 20+ years audio research; pioneering "creative weight" attribution — quantifying how much a work shapes AI outputs
Dataset Providers Alliance Alliance promoting ethical AI data sourcing; 5 compensation structures; cross-media (music, image, text, video)

Open-Source Tools

Library Stars Purpose License
librosa 8.2k Audio features, spectrograms ISC
essentia 3.4k Comprehensive MIR analysis AGPL-3.0
dejavu 6.7k Audio fingerprinting (Shazam-like) MIT
chromaprint 1.2k AcoustID fingerprinting LGPL
musicbrainzngs ~500 MusicBrainz API bindings ISC
Splink 4k+ Probabilistic record linkage MIT
MiRA New Music replication assessment Open
CLAP ~1k Language-audio embeddings Apache-2.0
Academic research highlights
Paper Year Key Contribution
Choi et al. — Large-Scale TDA via Unlearning (Sony) NeurIPS 2025 First large-scale training data attribution for music generation
Morreale et al. — Attribution-by-Design 2025 Inference-time provenance; argues TDA is impractical at scale
Barnett et al. — Musical Roots via Audio Embeddings 2024 CLMR and CLAP embeddings across 5M clips
Epple et al. — Watermarking Training Data 2024 Audio watermarks persist through training
Batlle-Roca et al. — MiRA ISMIR 2024 Model-independent replication detection at >10% rates
Afchar et al. — AI Music Detection ICASSP 2025 99.8% detection accuracy; poor robustness to manipulation

The two fundamental paradigms: post-hoc TDA (Sony — compute influence after training) vs. attribution-by-design (Morreale — embed provenance in architecture). This scaffold's A0–A3 levels bridge both: we don't require training-time access but incentivize it through higher assurance ratings.

Full survey: docs/planning/music-tech-landscape/04-academic-research.md

Market context
  • Market size: Generative AI in music projected $558M (2024) → $7.4B (2035) at 26.5% CAGR
  • Funding concentration: Suno ($375M+), Vermillio ($16M), Musical AI ($6M), >$250M equity across the sector in 2025
  • The uncomfortable truth (Water & Music): "Perfect attribution for music AI doesn't currently exist. All attribution approaches studied provide approximations at best."
  • This scaffold's position: To our knowledge, the only open-source, auditable, research-backed attribution infrastructure with formal uncertainty quantification

Full analysis: docs/planning/music-tech-landscape/


Technology Stack
Layer Choice Rationale
Language Python 3.13 Type hints, async, scientific ecosystem
Package manager uv 10–100x faster than pip, lockfile reproducibility
Web framework FastAPI + async SQLAlchemy Native async, auto OpenAPI docs, Pydantic integration
Database PostgreSQL + pgvector JSONB for metadata, vector similarity for embeddings
Agent PydanticAI Type-safe agent framework, FallbackModel for resilience
Agentic UI CopilotKit (AG-UI protocol) SSE streaming, tool-use visualization, state sync
Entity resolution Splink + jellyfish + thefuzz Fellegi-Sunter model, EM parameter estimation
MCP server MCP Python SDK (FastMCP) Machine-readable permission queries
Frontend Next.js 15, React 19, TypeScript strict App Router, server components
CSS Tailwind v4 + CSS custom properties Zero hardcoded hex, full design token system
State Jotai Atomic state for theme, role mode, works
Observability Prometheus + PostHog Metrics + product analytics
Quality ruff + mypy + pre-commit Fast linting, strict typing, 13 hooks

Each choice is documented as a decision node in the Probabilistic PRD. The PRD models 85 decision nodes across 5 levels, with conditional probabilities weighted by four team archetypes (Engineer-Heavy, Musician-First, Solo Hacker, Well-Funded).

Standards and Identifiers
Standard Body Purpose in Scaffold
ISRC IFPI Recording-level identifier (A1 assurance)
ISWC CISAC Musical work composition identifier (A2 assurance)
ISNI ISNI-IA Creator identity verification (A3 assurance)
IPI CISAC Interested Party Information — rights society ID
DDEX ERN DDEX Label/distributor metadata exchange (future)
C2PA C2PA Content provenance manifests (future)
ISO/IEC 42001 ISO AI management system compliance

Contributing

  1. Use uv for all dependency management — pip and conda are banned.
  2. Run pre-commit run --all-files before pushing. All 13 hooks must pass.
  3. Write tests for all new functionality. Follow tests/README.md.
  4. Type-annotate all public functions. Use from __future__ import annotations.
  5. Use pathlib.Path for file paths. Always specify encoding='utf-8'.
  6. Never modify sections marked with # AIDEV-IMMUTABLE.

See .claude/CLAUDE.md for the full development behavior contract.


Roadmap

Current focus areas tracked as GitHub Issues:

Area Status Issues
ETL orchestration (Prefect 3.x) Planned #26, #57
C2PA provenance manifests Planned #24
AG-UI production adapter In progress #30
OpenTelemetry instrumentation Planned #55
Bayesian network signal integration Planned #27
Voice/likeness protection (ELVIS Act) Planned #21
Feedback → confidence calibration loop In progress #20

References

Academic references used throughout the scaffold (30 citations)

Confidence and Calibration

  • Angelopoulos, A. N. & Bates, S. (2025). "Conformal Prediction: A Gentle Introduction." arXiv:2504.14154.
  • Quach, V. et al. (2024). "Conformal Language Modeling." arXiv:2512.12844.
  • Tian, K. et al. (2025). "SConU: Selective Conformal Uncertainty in LLMs." arXiv:2503.15850.
  • Beigi, H. et al. (2024). "Conformal Prediction for LLMs." arXiv:2410.20199.
  • Wang, Z. et al. (2025). "Uncertainty Quantification in LLMs." arXiv:2506.17419.
  • Stengel-Eskin, E. et al. (2025). "Calibrated Self-Rewarding Vision Language Models." arXiv:2601.15778.

Entity Resolution and Data Engineering

  • Fellegi, I. P. & Sunter, A. B. (1969). "A Theory for Record Linkage." JASA, 64(328), 1183–1210.
  • Papadakis, G. et al. (2021). "An Overview of Blocking Methods for Entity Resolution." ACM Computing Surveys.
  • Narayan, A. et al. (2024). "Can LLMs Replace Data Engineers?" arXiv:2512.23737.
  • Monarch, R. (2023). "An Overview of Human-in-the-Loop ML." arXiv:2307.15475.
  • Settles, B. (2009). "Active Learning Literature Survey." University of Wisconsin-Madison.

Attribution and Provenance

  • Teikari, P. (2026). "Governing Generative Music." SSRN No. 6109087.
  • Teikari, P. & Fuenmayor, E. (2026). "Digital Product Passports as Agentic Supply Chain Infrastructure." SSRN No. 6068907.
  • Morreale, F. et al. (2025). "Music and GenAI." arXiv:2510.08062.
  • Anthropic. (2025). "Model Context Protocol Specification."

Audio Security

  • Azzuni, M. & Saddik, A. (2025). "Audio Watermarking Survey." arXiv:2505.00579.
  • Li, Y. et al. (2025). "Music Watermarking." arXiv:2511.21577.
  • Nemecek, J. et al. (2025). "Adversarial Audio Watermark Removal." arXiv:2505.23814.

Concept Drift

  • Lu, J. et al. (2019). "Learning under Concept Drift: A Review." IEEE TKDE, 31(12).
  • Evidently AI. (2024). "ML Monitoring Documentation."

Standards and Governance

  • AICPA. (2017). "SOC 2 Trust Services Criteria."
  • ISO/IEC. (2023). "42001: AI Management System Standard."
  • IFPI. (2024). "ISRC Handbook."
  • CISAC. (2024). "ISWC Standard."
  • ISNI International Agency. (2024). "ISNI Technical Documentation."

Economics and Theory

  • Coase, R. H. (1937). "The Nature of the Firm." Economica, 4(16), 386–405.
  • Becker, G. S. (1968). "Crime and Punishment: An Economic Approach." Journal of Political Economy, 76(2), 169–217.
  • Posner, E. A. & Weyl, E. G. (2019). Radical Markets. Princeton University Press.
  • Bourdieu, P. (1984). Distinction: A Social Critique of the Judgement of Taste. Harvard University Press.

Development

This scaffold was developed with extensive use of Anthropic's Claude Opus 4.5 and Opus 4.6 via Claude Code, functioning as a pair-programming agent for multi-file refactors, test generation, and architectural reasoning. All figures were generated using Nano Banana Pro.


Citation

@article{teikariGoverningGenerativeMusic2026,
  title     = {Governing {Generative} {Music}: {Attribution} {Limits},
               {Platform} {Incentives}, and the {Future} of {Creator} {Income}},
  author    = {Teikari, Petteri},
  year      = {2026},
  journal   = {SSRN Electronic Journal},
  doi       = {10.2139/ssrn.6109087},
  url       = {https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6109087},
  keywords  = {Generative AI, Attribution Infrastructure, Music Industry,
               Platform Economics, Provenance Verification, Transaction Costs}
}

License

MIT | Detailed dependency licenses

from github.com/petteriTeikari/music-attribution-scaffold

Установка Music Attribution Scaffold

У этого сервера нет опубликованного пакета — он собирается из исходников. Открой репозиторий и следуй инструкции в README.

▸ github.com/petteriTeikari/music-attribution-scaffold

FAQ

Music Attribution Scaffold MCP бесплатный?

Да, Music Attribution Scaffold MCP бесплатный — установка в пару кликов через Unyly без оплаты.

Нужен ли API-ключ для Music Attribution Scaffold?

Нет, Music Attribution Scaffold работает без API-ключей и переменных окружения.

Music Attribution Scaffold — hosted или self-hosted?

Self-hosted: сервер запускается локально на твоей машине командой из раздела установки.

Как установить Music Attribution Scaffold в Claude Desktop, Claude Code или Cursor?

Открой Music Attribution Scaffold на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.

Похожие MCP

Compare Music Attribution Scaffold with

Не уверен что выбрать?

Найди свой стек за 60 секунд

Автор?

Embed-бейдж для README

Похожее

Все в категории development