ElevenLabs launches ElevenMusic as a licensed AI music platform, Splice ships generative AI tools that pay original sample creators, and Udio admits to YouTube stream-ripping as Sony's copyright case heats up.
ElevenLabs launches ElevenMusic — a licensed AI music platform for fans
ElevenMusic went live April 30 as a consumer-facing platform that combines streaming, remixing, and full song generation from text prompts. Users can shift tempo or genre on any track, regenerate individual sections via in-painting, or build entirely new songs from lyrics and mood descriptions. At launch it hosts roughly 4,000 artists — mostly emerging acts — with licensing deals through Kobalt and Merlin providing the legal backbone.
The free tier gives you 11 minutes of generation; the $9.99/month Pro plan unlocks up to 500 tracks, stem separation, and 500GB of storage. Commercial use requires stepping up to the Music Marketplace tier. If you're a working producer, the stem separation and in-painting are the genuinely useful bits — the rest is a consumer play competing directly with Suno and Udio on the creation side while adding a streaming layer on top.
Splice ships Variations and Craft — generative AI that pays sample creators
Splice launched two generative AI features in April: Variations (in the Splice Sounds plugin) generates alternate versions of samples — adjusting key, tempo, and structure while retaining the original character. Craft (in the INSTRUMENT plugin) transforms samples into playable instruments. A third tool, Magic Fit, which adapts any sample to match your session's harmonic and rhythmic context, ships this summer.
The meaningful innovation here is the compensation model: original creators get paid every time their sound is used as a source and every time a variation is downloaded. Every sample in Splice's 3M+ catalog remains traceable to its creator. If you already use Splice in your workflow, this is the first AI sample tool that doesn't create a compensation gap — it extends the existing pay-on-download system into generated outputs.
Udio admits to YouTube stream-ripping as Sony lawsuit escalates
In a filing reported May 1, Udio acknowledged using a YouTube stream-ripper to obtain training data — a significant admission in Sony Music's ongoing copyright case. Sony is the last major label still litigating against both Suno and Udio (UMG settled with Udio in October 2025; Warner settled with Suno in November 2025). The fair-use question at the center — whether training generative models on copyrighted recordings is transformative — is expected to produce a pivotal ruling this summer when the Suno case goes before Judge Denise Casper in July.
The stream-ripping admission adds a new dimension: it potentially invokes the DMCA's anti-circumvention provisions on top of standard copyright infringement claims. For anyone building on top of these platforms commercially, the legal ground remains unstable until summer rulings land.
Deezer: 44% of daily uploads are now AI-generated, 85% of their streams are fraud
Deezer published data showing 75,000 AI-generated tracks hit the platform daily — 44% of all new uploads. But here's the telling number: AI music accounts for less than 3% of total streams, and 85% of those streams are fraudulent bot activity that Deezer has demonetized. The platform is now licensing its AI-detection technology to the wider industry.
This confirms what practitioners already suspected: the AI music flood is largely a streaming fraud play, not a listener preference shift. Real humans aren't choosing this content. If you distribute AI-assisted work legitimately, the detection systems and platform policies being built now will matter to you — false positives are an inevitable side effect.
Luminate study: listener discomfort with AI music is growing
NPR reported on Luminate's "Generative AI in Entertainment 2026" study, which found consumer sentiment toward AI music dropped from net -13% to net -20% between May and November 2025. People are most uncomfortable with AI creating music that mimics existing artists' styles. About a third remain indifferent. The study suggests musicians speaking out against AI is actively moving public opinion, particularly among younger listeners.
For working audio professionals: this isn't about whether the tech works — it's about whether audiences will accept the label. Disclosure and authenticity framing matter more than they did a year ago.
Tamber raises $5M from Adobe Ventures for "sonic intelligence" production suite
LA-based Tamber, founded by musician-developer Zoe Wrenn, closed a $5M round backed by Adobe Ventures, Rackhouse VC, M13, and others ahead of a May launch. The platform uses text and voice prompts to search and transform a library built from real-world field recordings — not synthetic sources. Describe a mood, texture, or place and it returns matching sounds that can be adapted to your project.
The positioning is explicitly "assistive, not generative" — extending human workflows rather than replacing them. Worth watching if the library quality delivers, but we'll reserve judgment until the product is publicly available.
"Podslop" hits 39% of new podcast feeds
Bloomberg reported that over a recent 9-day span, 39% of approximately 11,000 new podcast feeds on the Podcast Index were likely AI-generated content. The term "podslop" — a nod to the AI slop problem in text and image — captures the scale of synthetic audio content now flooding distribution platforms. Unlike AI music fraud, podcast platforms haven't yet deployed systematic detection or demonetization tools.
Voicebox is an open-source, local-first AI voice studio that hit v0.5.0 on April 25 and has crossed 21K GitHub stars. It bundles voice cloning, text-to-speech across 7 engines (including Qwen3-TTS and Chatterbox), a multi-track timeline editor, real-time audio effects, and global dictation — all running entirely on your machine. Clone a voice from a few seconds of audio, generate speech in up to 23 languages, or give any MCP-aware AI agent a custom voice.
The practical appeal: no cloud dependency, no per-character billing, no data leaving your machine. GPU acceleration works on NVIDIA, AMD, Intel Arc, and Apple Neural Engine. If you've been paying ElevenLabs for voice work in post-production or prototyping, this is the first local alternative with comparable quality and serious workflow integration.
A clear pattern is forming: ElevenMusic launches with Kobalt and Merlin licensing, Splice's AI tools trace compensation back to creators, UMG co-launches a platform with Udio, and Warner cuts a deal with Suno. Meanwhile Sony holds the line in court and Deezer demonetizes 85% of unlicensed AI streams. The industry is quietly bifurcating into a "licensed generative" tier and a gray-market tier that platforms are learning to suppress.
For practitioners, this means the tools you choose now will determine whether your AI-assisted output lives in the legitimate ecosystem or gets caught in increasingly aggressive fraud filters. The summer Sony rulings will likely accelerate this split — either validating fair use (unlikely given the stream-ripping admissions) or establishing that licensing is the only viable path forward for commercial AI music generation.