Treble Lands $18M to Scale Voice Simulation for Robots
Paladin Capital Group leads an $18 million extension for Treble, whose physics-based acoustic models generate training audio for voice AI and robots.
In short
Treble, an Icelandic acoustics startup, has raised an $18 million Series A extension led by Paladin Capital Group to sell physics-simulated sound as training and test data for voice AI, wearables and robotics.
At a glance
- $18 million Series A extension, led by Paladin Capital Group.
- Also participating: KOMPAS VC, Frumtak Ventures, the EIC and Omega ehf.
- Treble raised $12 million in 2024; total funding now exceeds $40 million.
- Founded in 2020 by acoustic engineers Finnur Pind and Jesper Pedersen.
- Named customers include Amazon and Logitech.
Treble has closed an $18 million extension to its Series A. Paladin Capital Group led the round, with KOMPAS VC, Frumtak Ventures, the EIC and Omega ehf taking part. The Icelandic company has now raised more than $40 million in total, after a $12 million round in 2024.
Selling sound that was never recorded
Acoustic engineers Finnur Pind and Jesper Pedersen founded Treble in 2020 around one idea: model how sound physically travels through a space, then render the result as audio. The output is synthetic data for speech enhancement and noise suppression, plus evaluation runs that push a model through acoustic conditions a recording session would never cover.
Pind frames audio AI as a data problem rather than a modeling one. Almost everything in sound-related AI so far, he argues, has come from recordings and material scraped off the web, and accurate physics simulation offers another way to manufacture that data.
Hardware tested before it is built
A second business line points at devices. Manufacturers can rehearse speaker and headphone designs virtually, or check whether a smart speaker still catches a command when the person speaking stands somewhere else in the room. Smart glasses and AI wearables appear among the categories tested this way, and Amazon and Logitech are named as customers. Earlier this year, Treble and Hugging Face released a speech recognition benchmark built around realistic conditions.
Why robotics comes next
Beyond voice model developers, the customer list reaches into robotics, automotive and drone companies. Machines that move listen in surroundings that keep shifting, and recorded corpora rarely span that range. François Ruether of Paladin Capital Group described the platform in the same arc, from voice AI through wearables and robotics to physical AI.
What the report leaves open
Revenue, headcount and valuation are not disclosed, and the report does not say how the $18 million will be spent. Those are gaps in the record rather than zeros, and none of them can be inferred from the size of the round.
FAQ
How much has Treble raised in total?
More than $40 million, including a $12 million round in 2024 and the new $18 million Series A extension.
What does Treble's simulation platform actually do?
It simulates how sound propagates through a space and turns that into synthetic audio for training and evaluating voice AI models, and for testing audio hardware before a prototype exists.
Who uses Treble?
Amazon and Logitech are named as customers, with use cases spanning voice AI models, wearables, robotics, automotive and drones.
