Treble, an Iceland-based startup founded in 2020 by acoustic engineers Finnur Pind and Jesper Pedersen, raised $18 million in a Series A extension led by Paladin Capital Group. The round brings total funding to over $40 million, following a $12 million tranche in 2024. Existing investors KOMPAS VC, Frumtak Ventures, EIC and Omega ehf also participated.
The company's core business is a simulation platform that generates synthetic acoustic data and virtual prototyping environments for voice AI companies, robotics makers and consumer hardware manufacturers. Customers include Amazon and Logitech.
Treble's economics rest on a straightforward moat: physics-based audio simulation as an alternative to internet-scraped training data. Most speech AI has relied on recordings and web-harvested datasets. Pind argues that accurate physics simulation lets labs generate clean, labeled training data at scale—a meaningful edge as voice AI models demand increasingly diverse acoustic scenarios.
The company operates two main product lines. Its synthetic data platform handles speech enhancement, noise suppression and model training for voice AI teams. It also benchmarks voice models against realistic conditions; earlier this year Treble partnered with Hugging Face to launch a benchmark for speech recognition across varying acoustic environments.
The second vertical is hardware simulation. Treble collaborates with headphone and speaker makers to run virtual prototyping—testing how a device's microphone array or speaker placement affects audio processing. The platform can model how a smart speaker handles voice commands based on physical positioning. Recently it expanded to simulate voice interaction for smart glasses and other wearables.
Pind sees a near-term TAM in voice AI infrastructure—a crowded space with heavy investment from major cloud providers and startups—but the real opportunity is downstream: hardware makers scaling voice-first devices. As voice becomes the primary interface for smart glasses, wearable audio, and ambient AI, the demand for realistic acoustic simulation before manufacturing ramps should follow.
