Arts · OECD FoS 6.4 · China MoE 1352 音乐
Music
音乐
L2Local breakthroughs ↑ Rising Score 1.60 · Last reviewed 2026-10-22 · Next review 2027-01-19
BIGGEST BREAKTHROUGH 2026-10-20 AI · incrementalField · major
Spotify: 50M+ AI slop/spam tracks removed from recommendations this year, 75M+ removed outright; AI Personas excluded from recommendations by default ↗
AI: combines engagement, upload patterns and third-party AI detectors (acknowledged to be getting harder). Field: the largest streamer codifies human-first recommendation and labelling (Spotify’s figures)
Why this level
AI music moved from lawsuits to licensing: Warner settled with Suno and Udio, Universal with Udio, and Suno launched label-licensed models (2026); IFPI issued global principles for AI recordings on official charts.
S = 0.30·method + 0.35·discovery + 0.25·adoption + 0.10·education (education scores 0 when there is no evidence; if empty, the other weights are renormalised)
Sources
- latimes.com/entertainment-arts/business/story/2026-09-21/ai-music-licensing-deals-labels-artists-who-benefits
- ifpi.org/ifpi-rolls-out-global-principles-for-the-eligibility-of-recordings-developed-using-ai-in-official-music-charts-worldwide
Composite 1.5–2.49: at least one peer-reviewed/officially verified AI-key result, or one dimension at 3; routine workflow largely unchanged. All levels →
Research directions
Text-to-musicLicensed training & copyrightAI-assisted production
Papers
- MusicLM: Generating Music From Text
Text-to-music reaches usable quality
2023-01-26 · arXiv
- Jukebox: A Generative Model for Music
First model generating songs with vocals
2020-04-30 · arXiv
- Music Transformer
Transformers for long-form music
2018-09-12 · arXiv
- Heard But Not Said: Prompting Understates Temporal Information of Musical Moments in Frozen Large Audio-Language Models
Probes show frozen audio-language models encode when musical events occur, but prompting fails to extract it.
2026-10-21 · arXiv · Eunsik Shin
- Derivation Depth Exposes Reasoning Failures in Tool-Augmented Music Analysis
Varying derivation depth from MIR tool outputs reveals reasoning failures in tool-augmented music analysis.
2026-10-21 · arXiv · Rupali Rajesh, Sethuraman T, Savya Khosla, Srinidhi Sunkara
- MuLaCover: Flexible Cover Song Generation via Symbolic Tonal Control
Cover-song generation with symbolic tonal control and text-specified lyrics and style.
2026-10-19 · arXiv · Lynn Yi, Zihan Xiong, Junjie Cao, Xiaolong Weng
- MUB: An Identity-Masked Benchmark for Music Understanding in Multimodal Language Models
A 500-recording identity-masked benchmark of tempo, key, metre and structure.
2026-10-16 · arXiv · Kun Liu, Lyra Tao
- DuET: Note-Frame Dual Encoding for Piano Performance-to-Score Transcription
A dual-encoder transformer for piano performance-to-score transcription with beat tracking.
2026-10-15 · arXiv · Wei Zeng, Ye Wang
- MUSEBENCH: A Benchmark for AI-Generated Music Detection and Artifact Analysis
361K+ tracks for detecting AI-generated music from 10 models including Suno and Udio.
2026-10-14 · arXiv · Yi Chen Liu, Syed Irfan Ali Meerza, Jiawei Yu, Xi Guo
- PitchIn: Natural-Language Vocal Coaching from Unstructured Community Feedback
Audio-language models learn vocal coaching from unstructured singing-community comments.
2026-10-12 · arXiv · Hayeon Bang, Hounsu Kim, Joonyong Park, Jung-Jae Yu
- Harmonic Representations in Symbolic Music Transformers are Localized and Controllable
Harmonic representations in symbolic music transformers are localised and can steer generation.
2026-10-11 · arXiv · Adi Silberschein, Megan Wei, Tamar Rott Shaham
- From arXiv, bioRxiv, medRxiv and OpenAlex (last ~30 days); summaries written by AIMap from the abstract.
Milestones 9
Big glowing nodes are leaps: AI leap cool · Field leap warm · double ring = both
-
Industry
Spotify: 50M+ AI slop/spam tracks removed from recommendations this year, 75M+ removed outright; AI Personas excluded from recommendations by defaultAI: combines engagement, upload patterns and third-party AI detectors (acknowledged to be getting harder). Field: the largest streamer codifies human-first recommendation and labelling (Spotify’s figures)
Spotify NewsroomVendor claim -
Industry
LA Times: Suno builds a new model with major labelsCoverage of licensing deals
Los Angeles Times -
System
Suno releases new models built with Warner Music and BMG licencesFirst label-licensed commercial music models
Reuters -
Legal
Munich Regional Court rules Suno infringed GEMA-administered German song copyrightsAI: memorisation and recognisable outputs held to be reproduction. Field: one of the first major European losses for an AI music firm (appealable)
Billboard -
Policy
IFPI global principles for AI-developed recordings in official chartsFirst global chart rules for AI recordings
IFPI -
Industry
Warner Music Group settles with Suno and forms a licensing partnershipIndustry moves from lawsuits to licensing
Reuters -
System
MusicLM: generating music from text (Google)Text-to-music reaches usable quality
arXivPreprint -
System
Jukebox: generating raw-audio music with singing (OpenAI)First model generating songs with vocals
arXivPreprint -
System
Music Transformer: generating music with long-term structureTransformers for long-form music
arXivPreprint
No milestones match this filter.
More papers & repos 3
| Type | Title | Date |
|---|---|---|
| Paper | MusicLM: Generating Music From Text Preprint | 2023-01-26 |
| Paper | Jukebox: A Generative Model for Music Preprint | 2020-04-30 |
| Paper | Music Transformer Preprint | 2018-09-12 |
Latest news 11
- Show HN: Shazam but for AI Slop Musichn-show-aien
- ‘The only way to properly create value in music is to make lasting copyrights.’Music Business Worldwideen
- American jailed for commanding 10,000 bots to stream his own AI-generated songsTom's Hardware AIen
- Suno made a new AI music model with major labels. Here's what it meansLos Angeles Timesen