Mirelo, a Berlin-based artificial intelligence company that automatically generates sound effects for videos, has secured $41 million in seed funding.
The round saw investment from Index Ventures and Andreessen Horowitz (a16z), with participation from Atlantic.vc and TriplePoint Capital.
Founded in 2023, the startup develops foundation models that produce synchronized sound effects for video content. Mirelo says it addresses a gap in generative AI development.
While AI has advanced rapidly in text, image and video generation, audio technology has lagged behind, according to Mirelo. The startup says video creators currently spend hours searching stock libraries and manually syncing sound effects to visuals.
Mirelo claims to generate matching audio for videos “in a matter of seconds.”
CJ Simon-Gabriel, CEO and co-founder of Mirelo, said: “Think of the difference between talkies and silent films – video without sound has so much less feeling and atmosphere.”
“Our bigger mission is to become the audio layer for all visual content across videos, gaming, social media, films and beyond.”
CJ Simon-Gabriel, Mirelo
“Mirelo’s first step is about democratizing access, empowering everyone to create the sound that their (AI) video deserves. But we’ll also empower professionals to rework audio, to do more of what they love, to be more expressive and imaginative in what they can achieve, while handling the boring stuff such as synchronization. Our bigger mission is to become the audio layer for all visual content across videos, gaming, social media, films and beyond.”
The founders, Simon-Gabriel and CTO Florian Wenzel, met while working as AI researchers at Amazon Web Services Labs. Simon-Gabriel holds a PhD in machine learning and causal inference from the Max Planck Institute, where he studied under computer scientist Bernhard Schölkopf. He completed postdoctoral work at ETH Zurich, while Wenzel earned his PhD in deep learning from Humboldt University and previously worked at Google Brain.
Mirelo recently released Mirelo SFX v1.5, a video-to-sound-effect model accessible through an API and web application called Mirelo Studio. According to Mirelo, the models require “50 times less compute” than typical large language models while delivering “superior quality to any competitor so far according to external evaluations.”
Wenzel said: “There’s a deep affinity between music and engineering; maybe that’s why so many of Mirelo’s team are musicians, and why musicians have always been early adopters of new technology.”
“There’s something about the intersection of mathematical precision and expressiveness that seems to draw people to both fields.”
“There’s a deep affinity between music and engineering; maybe that’s why so many of Mirelo’s team are musicians, and why musicians have always been early adopters of new technology.”
Florian Wenzel, Mirelo
Georgia Stevenson, the partner at Index Ventures who led the investment, said: “Sound is too often an afterthought in video production, yet it’s what determines whether a video or game truly resonates with its audience. Mirelo gives creators a new form of expression, letting them move faster and sound better.”
“The team led by CJ and Florian combines cutting-edge AI expertise with an unparalleled focus on audio’s emotional power. It is a combination that positions them to reshape how the world experiences sound.”
“Sound is too often an afterthought in video production, yet it’s what determines whether a video or game truly resonates with its audience.”
Georgia Stevenson, Index Ventures
Index Ventures has also invested in other music tech firms, including futuristic instrument firm ROLI and concert discovery service Songkick. In 2019, it participated in a funding round for San Francisco-based crowdfunding membership platform Patreon.
Guido Appenzeller, partner at Andreessen Horowitz, also commented on the Mirelo investment, saying: “To date, a16z has invested in multiple world-leading generative models each with a different focus area. Mirelo is tackling one of the most technically challenging and least explored areas of generative media: a specialized model for sound effect creation.”
“CJ and Florian have assembled a research-driven team whose breakthroughs in tokenization, data curation, and conditioning rival far larger efforts and we’re excited to back Mirelo as they scale their technology for the next generation of video models.”
“Mirelo is tackling one of the most technically challenging and least explored areas of generative media: a specialized model for sound effect creation.”
Guido Appenzeller, Andreessen Horowitz
This marks a16z’s latest investment in the AI space. The venture capital firm has made investments in ChatGPT developer OpenAI, Elon Musk’s AI venture xAI, data and AI software infrastructure company Databricks, and AI audio startup ElevenLabs.
Last year, a16z said in a submission to the US Copyright Office (USCO) that training AI on copyrighted materials is fair use, and doesn’t amount to theft of intellectual property. The firm wrote: “When an AI model is trained on copyrighted works, the purpose is not to store any of the potentially copyrightable content (that is, the protectable expression) of any work on which it is trained. Rather, training algorithms are designed to use training data to extract facts and statistical patterns across a broad body of examples of content – i.e., information that is not copyrightable.”
Music Business Worldwide
