Google's Gemini Omni can now turn text into realistic video with minimal technical skill ^[S1], and Stable Audio 3.0 ships six-minute music tracks in ComfyUI ^[S2]. The models work. The question is no longer whether generative AI can produce professional-grade creative output—it's who gets to ship it, where, and under what terms.

That shift is clearest in two deals that redefine the creative stack. Spotify and Universal Music Group struck a revenue-sharing agreement allowing Premium subscribers to generate AI covers and remixes of licensed catalogues ^[S3]. Separately, Spotify partnered with ElevenLabs to offer AI audiobook narration for independent authors ^[S4]. Both moves hand distribution power to the platform, not the model lab. Spotify controls the subscriber relationship, the monetisation surface, and—crucially—the rights clearance layer that makes legal use possible at scale.

Meanwhile, Google declared itself a contender in AI design at I/O, unveiling tools aimed squarely at educators, SMBs, and general users ^[S5]. The pitch is access: Google owns Android, Search, and Workspace, so its creative agents sit inside the workflows users already inhabit. Model labs like Anthropic and OpenAI excel at frontier performance, but they lack direct consumer surfaces. Google's distribution advantage turns good-enough generation into a wedge that model-focused startups cannot easily counter.

The implication is that creative-tools value is migrating away from inference quality and toward platform lock-in and rights infrastructure. Generative models are converging in capability; the sustainable moats now belong to whoever owns the user, the licence, or the workflow integration. For investors, that suggests the next wave of creative-tools exits will come from companies that either control distribution or solve the permission problem, not those that chase another basis-point improvement in CLIP score.