Voice AI is exploding, but the models powering it have a dirty secret: they’re trained on messy, scraped internet audio. Enter Treble, an Iceland-based startup that believes physics-based simulation is the cleaner, smarter path forward. The company just closed an $18 million extension to its Series A, led by Paladin Capital Group. This fresh capital brings Treble’s total funding to over $40 million, validating its bet that synthetic data will outperform real-world recordings for next-gen audio tech.
Founded in 2020 by acoustic engineers Finnur Pind and Jesper Pedersen, Treble isn’t just building another dataset. It’s creating a virtual testing ground where hardware makers and AI labs can stress-test their products before they ever hit a factory line. Major players like Amazon and Logitech are already using the platform to prototype how speakers and headphones perform in complex acoustic environments. Instead of waiting for user complaints about poor noise cancellation, engineers can simulate a crowded restaurant or a windy street to tweak algorithms instantly.
The startup’s recent partnership with Hugging Face highlights its growing influence. Together, they launched a benchmark for speech recognition models that tests performance under realistic, simulated conditions rather than ideal studio settings. Pind argues that accurate physics simulation offers a scalable alternative to data scraping. “Audio AI is really a data challenge,” he noted, emphasizing that their approach allows for precise control over variables like distance, reverberation, and background noise.
Looking ahead, Treble is eyeing the booming market for smart glasses and hearables. Pind sees a future where these devices offer “superhuman hearing,” letting users mute distracting conversations or focus on voices within a two-meter radius. As robotics and automotive sectors begin integrating sound-based interactions, Treble’s simulation layer becomes critical infrastructure. It’s no longer just about recognizing words; it’s about understanding context in the physical world. If voice is the new interface, Treble wants to be the lab where it gets perfected.
