ElevenLabs’ new music-generation model can switch genres mid-track

📌 ElevenLabs releases Music v2 – genre‑switching and section‑level editing for AI‑generated songs

🎣 You thought AI music could only make short loops? ElevenLabs says you can now jump from opera to heavy metal in the same track and edit verses without touching the rest.

🤔 Why genre‑switching matters now
As AI‑generated music moves from demo clips to usable productions, creators need more control over style, structure and legal safety. Early models could generate melodies but struggled to keep coherence when the style changed or to let users edit only a part of a song.

🧪 How ElevenLabs built Music v2
The company trained the new model on licensed audio data that is cleared for commercial use, aiming to improve reliability across vocals, lyrics, arrangements and multiple languages. The release follows the first version launched roughly ten months earlier and arrives alongside similar pushes from Google, Stability AI and Suno, all of whom have recently unveiled models capable of longer, more complex tracks.

🔬 What Music v2 can do

• Switch genres in the middle of a track (e.g., opera → heavy metal → opera) without losing coherence.
• Generate fast rap verses while maintaining vocal quality.
• Insert non‑musical sound effects directly into the composition.
• Let users select a section of a song and regenerate it via prompts, leaving other parts untouched.
• Build a song by composing intro, verse, chorus, etc., as separate sections and then stitching them together.
• Deliver more consistent performance across different languages, lyrical content and vocal styles.

💡 Why this is useful for engineers and creators
Section‑level editing means developers can integrate the model into workflows where only a specific part of a track needs revision—think adaptive game soundtracks or personalized ad jingles. The commercially‑licensed output removes a major barrier for products that intend to distribute or sell the generated music, a point ElevenLabs stresses after seeing peers like Suno and Udio face copyright disputes.

⚠️ What we still don’t know
The article does not disclose the model’s size, training‑data volume, benchmark scores, or latency figures. Without those details it’s hard to compare Music v2’s efficiency or quality against competing offerings on a technical basis.

🎯 Takeaway for product teams
If you are building generative‑audio tools and need a model that can (a) change style mid‑song, (b) edit specific sections safely, and (c) provide royalty‑clear output, ElevenLabs’ Music v2 is now accessible via the ElevenCreative suite, the new ElevenMusic platform, and will soon be available through ElevenAPI.

🔗 Source
TechCrunch – “ElevenLabs’ new music‑generation model can switch genres mid‑track” by Ivan Mehta
https://techcrunch.com/2026/05/27/elevenlabss-new-music-generation-model-can-switch-genres-mid-track/

#AI #AIMusic #ElevenLabs #MusicGeneration #GenerativeAudio #TechCrunch #SoundDesign #MusicTech