IdeaWell

Project ideas from Hacker News discussions.

YuE2 · Frontier Music with Symbolic Planning

Original Article

Hacker News Discussion

📝 Discussion Summary (Click to expand)

1. AI music lacks soul and artistic depth
Many commenters feel that AI‑generated tracks are technically proficient but emotionally empty.

“That example ‘Comedy’ sounds like the most soulless synthetic example of ‘music’ ever. … It’s still technically impressive but it’s 0/10, absolutely empty of soul.” – philip-b
“All AI generated art feels somewhat masturbatory to me, but AI generated music is the most baffling of them all. It’s like what if someone thought Muzak had too much soul.” – slg

2. AI as a controllable tool for musicians
Several users see the model as a creative aid—like a “controlnet” for composition—allowing rapid iteration and augmentation.

“The use case is control. It’s essentially a controlnet for the actual music writing. So a musician could compose melody, harmony, and lyrics and generate many performances of the same song until they get one they like.” – popalchemist
“I’m a composer and I can 100% get behind this workflow… This is that.” – popalchemist
“I’ve been waiting for symbolic music generators for ages. This is so nice for musicians… generate variations on pieces, making pieces easier or harder.” – euleriancon

3. Worries about market saturation, discovery, and economic impact
A frequent concern is that floods of AI‑generated content will drown out human‑made music, devalue creative work, and make it harder for artists to earn a living.

“All of these projects are basically just destroying entire fields of human creativity.” – stevage
“It’s being handed an infinite slop machine, drowning out genuinely good content and making it harder than ever to discover. As it becomes harder to discover, fewer talented creatives are able to make a living.” – applfanboysbgon
“The majority of people… don’t want to listen to AI music … Systems like this will produce muzak for ubers, cafes, phone queues, advertisements.” – bit‑rot

🚀 Project Ideas

HumanScore

Summary

  • A browser extension and web app that labels tracks on streaming platforms (Spotify, YouTube, SoundCloud) as likely human‑made or AI‑generated using audio fingerprinting, metadata analysis, and community voting.
  • Core value: restores discoverability for listeners who want authentic human music amid the flood of AI‑generated slop.

Details

Key Value
Target Audience Music listeners, curators, playlist creators frustrated by AI‑generated content drowning out human art
Core Feature Real‑time authenticity scoring overlay on music platforms, with optional filters to hide low‑scoring (likely AI) tracks
Tech Stack JavaScript/TypeScript (extension), Python backend (audio analysis with librosa, essentia), PostgreSQL, React for web dashboard, optional ML model for AI‑vs‑human detection
Difficulty Medium
Monetization Revenue‑ready: Freemium – free basic labeling, premium API access for curators and playlist services ($9/mo)

Notes

  • HN users complained: “It becomes harder to find and therefore listen to music you like when every avenue of discovering music is drowned out with 1000 AI‑generated songs for every 1 human‑made song.” (stevage) and “How about… figure out how to get a few dozen of them motivated to work on that instead of sitting and moping for the rest of our lives about the end of human creativity?” (mindcandy). HumanScore directly tackles this discoverability pain.
  • Provides a practical utility: listeners can regain trust in platforms, curators can build human‑focused playlists, and musicians gain a better chance to be heard.

MuseControl

Summary

  • A DAW‑compatible plugin (VST/AU) and standalone app that lets musicians import symbolic music (MIDI, ABC, MusicXML) and use a controllable generative model (like YuE2) to produce multiple performances, adjust genre, lyrics, instrumentation, and export individual stems (vocals, backing, etc.).
  • Core value: gives musicians fine‑grained, iterative control over AI‑assisted music creation, turning the model into a “controlnet for music” rather than a black‑box slop generator.

Details

Key Value
Target Audience Composers, producers, songwriters who want AI as a tool for experimentation and rapid iteration
Core Feature Symbolic‑to‑audio generation with adjustable parameters (style, lyric, arrangement) and stem‑level export, integrated into DAW workflow
Tech Stack C++/JUCE for plugin, Python torch model (YuE2 or similar), ONNX runtime for real‑time inference, JSON preset system, optional WebAssembly version for browser demo
Difficulty High
Monetization Revenue‑ready: Subscription ($14.99/mo) or perpetual license ($129) with free trial

Notes

  • Popalchemist noted: “It's essentially a controlnet for the actual music writing… generate many performances of the same song until they get one they like.” and euleriancon praised: “Being able to intelligently translate parts, generate additional parts, generate sheet music from audio, generate variations on pieces…”. MuseControl directly enables that workflow.
  • Addresses the frustration of AI music feeling soulless by placing the human in the loop for creative decisions, letting artists steer the output toward their taste.

VeriTune

Summary

  • A lightweight provenance service where musicians can register their creations (audio file + metadata) with a cryptographic signature or NFT‑like token, generating a verifiable “Human‑Made” badge that platforms can display.
  • Core value: combats AI‑generated misattribution and gives listeners a trustworthy way to support genuine human creators.

Details

Key Value
Target Audience Independent musicians, labels, and platforms wanting to prove human authorship
Core Feature Upload‑and‑sign workflow producing a tamper‑proof attestation (via Ethereum/EIP‑712 or decentralized ID) that can be queried by music services to show a verified badge
Tech Stack Node.js/Express backend, IPFS/Filecoin for storage, Ethereum L2 (Polygon) for low‑cost signatures, React verification widget, optional OAuth integration with streaming services
Difficulty Medium
Monetization Revenue‑ready: Pay‑per‑registration ($0.99) or subscription for bulk uploads ($19/mo for 100 tracks)

Notes

  • Users expressed distrust: “I have zero interest in it if it's AI‑generated.” (hypersoar) and “It’s being handed an infinite slop machine… making it harder than ever to discover.” (applfanboysbgon). VeriTune gives a concrete way to prove human origin, addressing the core anxiety.
  • Enables discussion around digital provenance, incentivizes platforms to adopt verified‑human labels, and could revive confidence in music markets flooded with AI slop.

Read Later