In short
- ElevenLabs launched Music v2, able to switching genres mid-track, constructing songs part by part, and inpainting particular components.
- Stability AI launched Steady Audio 3.0, a four-model household with open weights for 3 variants, skilled on licensed knowledge, producing tracks as much as six minutes and twenty seconds lengthy.
- Each releases lean exhausting into licensed coaching knowledge—however Suno, valued at $2.45 billion with roughly 100 million customers, remains to be the platform most individuals attain for first.
Two vital AI music updates landed this week, and neither got here from Suno.
ElevenLabs, the Polish-founded voice AI firm sitting at an $11 billion valuation after a $500 million Collection D in February, launched Music v2. Stability AI—the Steady Diffusion folks—dropped Steady Audio 3.0, a four-model household with open weights and tracks that run previous six minutes.
The backdrop is the Recording Business Affiliation of America copyright fits from 2024 towards Suno and Udio, which made “skilled on licensed knowledge” a very powerful phrase in any AI music announcement. Each ElevenLabs and Stability are leaning on that onerous, ensuring you received’t have points with the outputs you generate.
Music v2: One observe, opera to heavy metallic, no breakdown
Music v2 is ElevenLabs’ second music mannequin, arriving roughly 10 months after the primary. The core pitch is coherence below strain. In line with Elevenlabs, a single observe can shift from opera to heavy metallic and again, maintain collectively by quick rap, and embed non-musical sound results—all with out the composition coming aside.
Generative audio tends to disintegrate precisely when prompts get difficult, so that is the factor price watching, particularly in longer compositions.
Inpainting is now truly helpful: choose a piece, regenerate it, go away all the things else untouched. Customers also can construct songs part by part—intro, verse, refrain—with the mannequin sustaining continuity all through as an alternative of treating every clip as a standalone technology. Multilingual help has improved too, although ElevenLabs did not publish specifics.
The mannequin powers three platforms: ElevenMusic for creators, ElevenAPI for builders, and ElevenCreative for manufacturers. It is stay on ElevenMusic and ElevenCreative now; API entry is early-entry through the gross sales crew.
ElevenLabs additionally lower Music v1 and v2 pricing by as much as 50% for ElevenAPI and as much as 40% for ElevenCreative self-serve. The corporate hit $500 million in annual recurring income in April 2026. Music remains to be a small slice of that—however ElevenMusic, which launched as a shopper app in April, is a direct shot at Suno’s consumer base.
Steady Audio 3.0: Open weights, on-device, truly longer
Steady Audio 2.0 topped out at three minutes and was already behind Suno when it launched in 2024. Steady Audio 3.0 ships 4 fashions: Small SFX (on-device sound results), Small (full music composition on-device), Medium (as much as 6:20, stronger {hardware}), and Massive (API-only). Three of the 4 have open weights on Hugging Face.
The Small fashions run at 459 million parameters every—no GPU wanted. (Parameters are what measure an AI mannequin’s capability, basically.) Medium hits 1.4 billion parameters and generates its 6:20 output in about 1.31 seconds on an H200 GPU. Massive, at 2.7 billion, is API-only for organizations with over $1 million in income. Per-second technology granularity means you get precisely the observe size you requested for, not an approximation.
It’s additionally supported in ComfyUI for native setups
The structure is new: a semantic-acoustic autoencoder Stability calls SAME, designed to carry melodic coherence over longer outputs. LoRA fine-tuning is supported, so artists can adapt the fashions to their very own catalogs. Inpainting is in too—single-segment, multi-segment, and causal continuation to increase a observe previous its authentic endpoint.
For context, a LoRA (Low-Rank Adaptation mannequin) is sort of a tiny mannequin that situations how the total mannequin generates its outputs. For those who prepare a LoRA on blues, the mannequin will produce blues, in the event you prepare a LoRA on BB King blues, the mannequin will produce songs that may sound like BB King. Inpainting means a mannequin can repair small errors in its creation. So, for instance, if the mannequin hallucinates one thing on the 2:30 mark, you’ll be able to choose a number of seconds of the music, ask the mannequin to alter it into no matter you need, and the mannequin will generate a chunk of the music that matches completely in that timeframe and blends with the precise music as a complete.
Stability has been technically credible in AI music for years with out breaking by commercially. The open-weight play is the Steady Diffusion technique utilized to audio—seed the developer group, see what will get constructed. The licensing is cleaner than something Steady Audio has shipped earlier than, with partnerships in place with Common Music Group and Warner Music Group.
The goal: Suno, the AI music king
If ChatGPT is the king of AI textual content, Suno is the king of AI music. The corporate behind the mannequin hit a $2.45 billion valuation in November 2025, crossed $300 million in annual recurring income, and has been utilized by roughly 100 million folks.
It generates round 7 million songs per day. Warner Music settled its go well with towards Suno in November 2025; Sony and UMG are nonetheless in federal courtroom.
To keep away from these copyright wars, ElevenLabs has licensing offers with Consider, Kobalt, and Merlin. Stability has Warner and Common. Udio settled with all three majors and is now a walled backyard—nothing you generate can go away the platform.
Steady Audio 3.0 Small and Medium can be found on Hugging Face now. Massive is stay through the Stability AI API. Music v2 is free for ElevenMusic customers, with industrial tiers by ElevenCreative and ElevenAPI.
Day by day Debrief Publication
Begin every single day with the highest information tales proper now, plus authentic options, a podcast, movies and extra.