The Napster co-founder and CEO Prem Akkaraju are repositioning the maker of Stable Diffusion away from images and toward music — three new audio models, label-licensed training data, and a plan to let you steer generation by humming or beatboxing.
2026-10-02 · by Moe Ameen
Stability AI, the company best known for the Stable Diffusion image models, is being rebuilt around music. In reporting published October 2, 2026 — Sean Parker told The Information, in an interview recapped by TechCrunch — the Napster co-founder and CEO Prem Akkaraju described a strategic shift to position Stability as an AI toolmaker for professional musicians rather than an image-generation company. Parker joined an $80 million rescue of Stability about two years earlier, after the startup's overspending and internal turmoil led to the departure of founder Emad Mostaque; he and Akkaraju have been reshaping its direction since.
The music push is anchored by the Series B that Stability announced in late August 2026: $76 million backed by the three major record labels — Universal Music Group, Sony Music Group, and Warner Music Group — along with Electronic Arts and others. As part of that deal the labels licensed their catalogs to Stability for training, which is the piece that makes a "built for professionals" pitch credible — the models are being trained on cleared, licensed music rather than scraped audio.
Since the round, Stability has released three new audio models and AI music-editing software. The tools can generate full instrumental tracks or shorter musical snippets from a text prompt. Parker said an upcoming update will let users steer generation in more musical ways — humming a melody or beatboxing a drum pattern to guide the output — though no release date has been given, so that remains a plan rather than a shipping feature.
Parker also framed the rebuild as a change in posture. He has reportedly conceded that his old instinct — asking for forgiveness rather than permission — did not work out well the last time around, and is said to be playing by the rules this time. That is a notable stance given that Stability has been a central target in the AI-copyright fight, and that the labels now backing it spent the prior two years litigating generative AI.
This one is about audio, so the Kompozy bridge is specific: a licensed AI track fixes what music you can legally use — it does nothing about the video that music plays under. Sound-on feeds punish the wrong audio (copyright strikes, muted clips, suppressed reach), so a cleared instrumental from Stability's new tools is a real unlock. But a 30-second bed is not a post. [Kompozy](/) is where the post gets made and shipped: feed it a source — a talk, a demo reel, a transcript — and it generates the actual video, not just the sound layer. [Marketing Shorts](/glossary/output-buckets) composite a hook, footage, and a music track into a finished vertical clip; [Clipped Shorts](/glossary/content-repurposing) cut long footage into several captioned 9:16 cuts; [Persona Shorts](/glossary/persona-shorts) put a face-locked avatar on-camera to deliver the message. Score the bed in Stability, drop it under the clip Kompozy built, and you have a scroll-stopping short with legal audio instead of a silent one.
The second half is distribution, which is where a single soundtrack-plus-clip has to become a week of posts. Kompozy fans one source into the formats a music model will never make — brand-exact [Carousels](/glossary/hyperframes), quote graphics, blog articles, and email newsletters — all held to one voice by a [Persona Brief](/glossary/persona-brief), then [Autopilot](/glossary/autopilot) schedules and publishes the set across the eight social platforms plus blog and email behind a per-post review gate. And because Kompozy treats every generator as an interchangeable input, whatever Stability ships next from this music push — the humming-steered update, a stronger audio model — slots in as a better soundtrack source without you rebuilding the pipeline around it.
Under co-founder-turned-backer Sean Parker and CEO Prem Akkaraju, Stability AI is being repositioned from an image-generation company (Stable Diffusion) into an AI toolmaker for professional musicians. It has released three new audio models and AI music-editing software that can generate full instrumental tracks or short snippets from text prompts, with a planned update to let users steer generation by humming or beatboxing.
Yes. As part of the $76 million round announced in late August 2026, Universal Music Group, Sony Music Group, and Warner Music Group licensed their catalogs to Stability AI for training. That licensed foundation is central to the "built for professionals" pitch and is what makes the generated music lower-risk to use than audio from models trained on scraped songs.
Not yet. Sean Parker described humming a melody or beatboxing a drum pattern to steer generation as an upcoming update, but no release date has been announced — so it is a plan, not a feature available today. The current tools generate from text prompts.
A generated track is a soundtrack, not a finished post. The practical workflow is to score a cleared instrumental or hook with a tool like Stability's, then build and publish the actual video with a content engine like Kompozy — which composites the clip, burns captions, reframes it per platform, and schedules it across the eight social platforms plus blog and email.