Stem Player founder Alex Klein on reverse-engineering stems from any song, interactive streaming deals, and why generative music sounds like Muzak

Summary

Alex Klein, founder of Stem Player, has quietly built what he argues is the most sophisticated music intelligence platform available to consumers, securing signed interactive streaming deals with major labels for full catalog access on Stem FM — a milestone that has eluded other AI music ventures.

The Technology

Stem Player's core capability is source separation: a discriminative AI approach that reverse-engineers vocals, bass, drums, and instrumentals from a standard mixed audio file, no stems upload required. Klein frames stems as the musical equivalent of tokens in a large language model — the atomic unit that enables next-token-style prediction and recombination. The platform demonstrated live mixing of Kehlani, Lil Yachty, Kendrick Lamar, LeCouve, and Mac Miller drums in real time during the segment.

Beyond separation, the team built an entirely new audio codec called Stem FM format, built around a concept Klein calls music-aware processing. Unlike Spotify's infrastructure, which treats audio as undifferentiated frequency bands, Stem FM's codec is beat- and key-aware at the chunk level, effectively encoding DJ-grade intelligence into the streaming layer itself. Klein calls it the hardest engineering problem the team has ever solved.

Stem Player also maintains a three-year research partnership with Queen Mary University's Centre for Digital Music, with submitted and accepted papers on beat structure and segment detection in music information retrieval.

Licensing and Business Model

The company has signed interactive streaming deals with the major labels covering full catalog access — a meaningful structural advantage over competitors who have faced legal resistance. Stem FM commits to no advertising for at least the first two contract terms, a condition baked into the label agreements.

On the revenue model, Klein is proposing a departure from per-stream payouts with a system called Time-Based Artist Compensation, or TBax. Under TBax, a subscriber's monthly fee is allocated to rights holders in proportion to actual listening time spent with each artist. A $20 monthly subscriber who spends half their listening time with one artist would direct $10 of that fee to that artist's label. Klein positions this as a fix to what he calls the structural dysfunction of per-stream economics, which he traces back to CD-era mechanical licensing logic.

Hardware and Consumer Positioning

The current Stem Player device is priced at $299, positioned as a consumer product — not prosumer — and described as the loudest portable Bluetooth speaker in its class with onboard compute and storage. The company previously sold out an earlier version, generating approximately $20 million in revenue. The device has been associated with artists including Kanye West (an early distribution partner, now unaffiliated for four years), Ghostface Killah, Quavo, Justin Bieber, Metro Boomin, and Tsmino. A first-of-a-thousand-units founder's edition is available at stemplayer.com.

Klein founded the broader company in 2013 in London, initially building programmable computers for children sold as modular kits. That business generated over $100 million in revenue at retail before he pivoted to music.

On Generative AI Music

Klein is sharply critical of text-to-audio platforms. He attributes the poor audio quality of outputs from Suno and Udio to a fundamentally flawed paradigm — diffusing audio frame by frame — and compares the results to Muzak. He also notes that in legal filings tied to the combined $5 billion-plus damages claim being pursued by every major label and publisher against Suno and Udio, statements made by those companies about their products' behavior were, in his view, demonstrably false — including claims that users could not prompt outputs using artist names or recognizable melodies like Smoke on the Water.

Klein's argument is that recombining real stems from existing recordings — the approach Kanye West pioneered through sampling — produces music that sounds and feels better than anything generated from scratch. Artists, per his account of a conversation with Smino (3.5 million monthly Spotify listeners), are not actually using generative audio tools in their creative process despite producers pitching them.