Treble, a Reykjavik-based startup founded by acoustic engineers, raised $18M in an extension of its Series A led by Paladin Capital Group.
KOMPAS VC, Frumtak Ventures, EIC and Omega ehf participated. The company took in $12M in 2024, bringing its total raised to more than $40M. Amazon and Logitech are customers.
Its product is simulation, not recording. Labs use the platform to manufacture synthetic audio for speech enhancement, noise suppression and training runs, and to grade models across conditions so results can be fed back to developers. A partnership with Hugging Face this year produced a benchmark scoring speech recognition models in realistic environments.
Pind’s argument starts with where training audio comes from. Almost every sound model so far has learned from real recordings or material scraped off the web, he told TechCrunch, which makes data supply the field’s real constraint. Physics simulation, in his view, is an alternative source – and the opening for both next-generation models and the hardware they run on.
Treble also works with headphone, speaker and robotics companies on virtual prototyping, letting them understand how a product will sound before building it.
The bet is that as voice becomes a primary interface for AI agents, smart glasses and robots, whoever controls the test environment shapes how those products perform in the real world.
