ElevenLabs launched Music v2 with mid-track genre switching and halved API pricing, Stability AI released open-weight Stable Audio 3.0 models generating six-minute compositions on consumer hardware, Spotify and UMG struck a landmark licensing deal for fan-made AI covers, and Sony dramatically expanded its copyright suits against both Suno and Udio.

ElevenLabs Ships Music v2 With Mid-Track Genre Switching and Section Editing

ElevenLabs released Music v2 on May 26, a significant upgrade to its music generation model. The headline feature is mid-track genre switching — the model can sustain fast rap, pivot to opera, and shift into heavy metal within a single track while maintaining musical coherence. Section-by-section editing lets you regenerate just a bridge or chorus without touching the rest of the song. Multilingual lyrics now work reliably across several languages, and vocal realism got a noticeable bump.

Pricing got more interesting: ElevenLabs cut API costs by up to 50% (now $0.15/minute of generation) and ElevenCreative pricing by up to 40%. All outputs are cleared for commercial use, trained exclusively on licensed data — a pointed distinction from Suno and Udio, both still fighting major-label copyright suits. If you produce music for ads, branded content, or apps and need commercially safe generation with granular editing control, this is the most capable option right now.

Source: TechCrunch, The AI Insider, ElevenLabs Blog

Stability AI Releases Open-Weight Stable Audio 3.0 — Six-Minute Songs, No GPU Required

Stability AI shipped Stable Audio 3.0 on May 20: a family of four models (459M to 2.7B parameters) that generate music up to 6 minutes and 20 seconds — more than double the previous version's ceiling. Three models ship with open weights on Hugging Face: Small SFX, Small, and Medium. The Large model (2.7B) is API-only, with enterprise licensing required for companies above $1M revenue.

The practical breakthrough is speed and accessibility. The Medium model generates audio in under two seconds on an H200 GPU and runs in seconds on a MacBook Pro M4. The Small models run on CPU only — no GPU needed. All models are trained on fully licensed and Creative Commons data, so outputs are commercially clear under the Stability AI Community License. If you build audio tools or need local music generation without copyright anxiety, this is the most capable open-weight option available.

Source: TechCrunch, Digital Music News, Music Ally

Google Flow Music Gets AI Agents, Mobile App, and Gemini Omni at I/O

Google used I/O 2026 on May 28 to push Flow Music (formerly ProducerAI, formerly Riffusion) well beyond its text-to-music origins. The biggest addition is Flow Agent — an AI creative assistant powered by Gemini that can plan and reason through complex musical tasks while remaining under the creator's control. Section-by-section editing now lets you restyle or translate only part of a track. Full-track covers and music video creation arrive via Gemini Omni Flash.

Flow Music also got its first dedicated mobile app (iOS now, Android soon). The platform runs on Lyria 3 Pro, generates full-length songs for free, and Google does not claim ownership of outputs. If you're a Believe or TuneCore artist, Flow Music is now integrated directly into your distribution workflow through a partnership announced earlier this month. Google is clearly positioning this as the free, full-featured alternative to Suno and Udio — and the agent-based workflow is genuinely different from anything the competitors offer.

Source: Google Blog, MusicTech, Android Central

Spotify and UMG Strike Landmark Deal for Licensed AI Covers and Remixes

Spotify and Universal Music Group announced licensing agreements on May 21 that will let fans create AI-generated covers and remixes of songs from participating UMG artists. The tool launches as a paid Spotify Premium add-on with a revenue-share model built on three pillars: consent, credit, and compensation. Artists opt in — nobody's catalogue gets remixed without permission.

This is the first major streaming platform to formalize AI-generated derivative works with label backing. Spotify is in talks with Sony, Warner, Merlin, and Believe for similar deals. No pricing or launch date yet, but the framework matters more than the timeline — it's a licensing template that could become industry standard. For producers and sound designers, the implication is clear: the market for AI-assisted remix and cover tools just got legitimate distribution infrastructure.

Source: Spotify Newsroom, TechCrunch, DJ Mag

Sony and UMG Escalate Copyright Suits Against Suno and Udio by Orders of Magnitude

Two filings this week dramatically raised the legal stakes for AI music generation. On May 25, Sony Music filed to expand its Udio lawsuit to cover 30,000 allegedly infringed recordings. The next day, UMG and Sony jointly moved to add over 61,000 recordings to their Suno case in Massachusetts — roughly a 100x increase from the original complaint's scope. The labels used audio fingerprinting and targeted prompts to identify their works in training data after Suno reportedly refused to disclose sources during discovery.

Settlement talks between UMG and Suno hit an impasse in April, with the labels demanding equity stakes and higher per-stream royalties. Warner settled with Suno last November; Sony remains the last major label actively litigating both platforms. A fair-use ruling is expected this summer and would set binding precedent for every AI music company. If you use Suno or Udio in production workflows, the legal ground under those tools is still shifting.

Source: Digital Music News, Music Business Worldwide, Yahoo Finance
Tool of the week
Stable Audio 3.0 Medium (Open Weight)

The open-weight Medium model (1.4B parameters) is the sweet spot of the Stable Audio 3.0 family. It generates full compositions up to 6:20, runs in seconds on a MacBook Pro M4, and ships with weights on Hugging Face and training/inference code on GitHub. All training data is fully licensed, so outputs are commercially clear under the community license.

Why it matters: this is the first open-weight model that can generate production-quality, full-length music locally on consumer hardware. If you build audio products, need offline generation for installations or live performance, or want to fine-tune a music model on your own catalogue, this is the practical starting point. The Small model even runs on CPU — useful for embedded or mobile applications where GPU access isn't guaranteed.

Open weights free (community license) · Enterprise license for >$1M revenue · API available for Large model · stability.ai/stable-audio
Worth watching
AI music's two-track future is crystallizing fast

This week made the split unmistakable. On one track: licensed-data platforms racing to build legitimate infrastructure. ElevenLabs and Stable Audio train on cleared catalogues and compete on quality. Google gives away Flow Music and partners with distributors. Spotify builds licensed remix tools with UMG revenue-sharing. On the other: Suno and Udio face ballooning lawsuits and stalled negotiations, even as Suno crosses $300M ARR and 2M subscribers.

The practical question for anyone building with AI audio: which track are you on? Tools trained on licensed data now match or exceed the unlicensed incumbents in quality and duration. The legal risk premium of using Suno or Udio in commercial workflows is rising weekly. If you haven't audited your AI music pipeline for copyright exposure, this is the week to start.