ElevenLabs launches ElevenMusic, Google ships Lyria 3 Pro with 3-minute structured tracks, Suno rewrites its ownership rules, and the UK kills its AI copyright opt-out plan.
ElevenLabs enters the music race with ElevenMusic
ElevenLabs — the voice AI company valued at $11B — launched ElevenMusic, an iOS app for AI song generation that puts it in direct competition with Suno and Udio. Free tier gives you 7 songs/day via text prompts; Pro ($9.99/mo) unlocks 500 tracks/month with expanded styles. It also includes Spotify-style discovery features like live stations, mood mixes, and remixing of other users' tracks. Worth testing if you need quick background music or scratch tracks, though it's consumer-facing rather than production-oriented.
Google ships Lyria 3 Pro — 3-minute structured tracks via API
Google's Lyria 3 Pro generates tracks up to 3 minutes (up from 30 seconds with Lyria 3) and understands song structure — you can specify intros, verses, choruses, and bridges. Available in Vertex AI, Google AI Studio, the Gemini API, and ProducerAI. All output is watermarked with SynthID. For anyone building tools or pipelines that need programmatic music generation, this is the most capable API-accessible model right now.
Suno rewrites its rules post-Warner settlement
Following its landmark deal with Warner Music Group, Suno is quietly tightening everything. Users no longer "own" generated songs — even paid subscribers get commercial use rights but are "generally not considered the owner." Free tier songs are no longer downloadable, only streamable. Paid tiers get monthly download caps. New models in 2026 will train exclusively on licensed material. With 2M paid subscribers and $300M in annual revenue, Suno is becoming more like a label-controlled platform than an open creative tool. Udio remains a walled garden with no downloads at all.
UK scraps AI copyright opt-out after massive backlash
The UK government officially abandoned its plan to let AI companies train on copyrighted music without permission. The original proposal would have made unauthorized training the default, requiring artists to opt out. Over 10,000 consultation submissions — 97% opposed — killed it. The government will now pursue voluntary licensing codes and transparency obligations instead. A significant win for creators, and a signal that the "train first, ask later" era may be closing in major markets.
Amorph gets a major visual update — text-to-plugin keeps evolving
Amorph, the free text-to-DSP plugin from Artists in DSP, received a significant GUI update this week. The plugin lets you describe an effect or instrument in plain English, uses an LLM to generate Cmajor code, and compiles it into a working VST3/AU/CLAP plugin inside your DAW. It's two plugins in one (FX + instrument) and it's genuinely unlike anything else in the plugin ecosystem. Still in open beta, still free.
LALAL.AI ships API v1 with multi-stem separation and voice cloning
LALAL.AI released its production-ready API v1, letting developers embed studio-grade stem separation (vocals, drums, bass, guitars, piano, strings, winds — all in a single request) and consent-based voice cloning into their own apps. Comes with OpenAPI spec and Swagger docs. If you're building audio tools or pipelines and need programmatic stem separation without training your own models, this is the most straightforward commercial option available.
What it is: A plugin that turns text descriptions into working audio effects and instruments inside your DAW. Describe "a granular delay with pitch-shifting feedback" in plain English, let an LLM generate the code, paste it in, and Amorph compiles it into a playable plugin with auto-generated knobs.
Why it matters: This isn't AI generating audio — it's AI generating the tools that generate audio. That's a fundamentally different proposition. Instead of replacing your creative decisions, it removes the barrier between imagining a sound processor and having one. Power users can tweak the Cmajor code directly; everyone else just types.
The Suno/Warner deal is the canary in the coal mine. AI music platforms are quietly shifting from "you own what you create" to "you license what we generate." Suno's new terms explicitly say users are "generally not considered the owner." Udio won't even let you download. Meanwhile, the UMG v. Suno fair use ruling expected this summer could reshape the entire legal landscape.
If you're using AI-generated music commercially — in sync, games, content — the licensing ground is shifting under your feet. Read the terms of service now, not after you've shipped something built on output you don't actually own.