← Portfolio · Portfolio case · AI infrastructure, generative audio · Germany
Mirelo: the missing audio layer for generative video
Seed investor from day one · mirelo.ai · All companies
Mirelo builds foundational music and sound models. Its starting point is a fundamental gap in generative media: AI video without sound is incomplete. By grounding state-of-the-art research in a deep understanding of music, audio and creativity, the team is building the missing audio layer for next-generation video, games and interactive media.
Why we invested
ellipsis has been an investor from day one. From the first meeting it was clear that Carl Johann Simon-Gabriel, Florian Wenzel and the team are the kind of world-class, category-defining founders we want to partner with: research depth in audio combined with the product sense to ship models people can actually work with. The moat is expertise and data: models grounded in how sound designers and musicians work, and a growing corpus from real production use.
What we did
We worked closely with the founders on product, hiring and fundraising through to their Series A.
Milestones
- October 2025. SFX 1.5, its fastest video-to-sound model, on studio.mirelo.ai, Replicate and fal (1.7x faster generation).
- December 2025. $41M Series A led by Index Ventures and Andreessen Horowitz, a highly competitive round.
- May 2026. SFX 1.6: from generation to iterative sound editing (long generations, extender, ambiancer, inpainter), the first model that treats sound as something you iterate on.
- Summer 2026. A Speech-Preservation API that regenerates ambient sound without killing the dialogue underneath, a major Audio-to-MIDI update, and MuScriptor, released with Kyutai.
Milestones are compiled from the company's own announcements and our Dispatch newsletter; figures are the company's.