Treble Technologies Raises $18M to Add Hearing to Robot Perception

Robotic perception has recently focused on vision and touch, leaving hearing relatively quiet. Treble Technologies is changing that by turning sound propagation in physical spaces into simulatable, generative data.

Founded in 2020 by Finnur Pind and Jesper Pedersen in Reykjavik, Iceland, Treble has raised $18 million in an A-2 round led by Paladin Capital Group, with participation from existing investors KOMPAS VC, Frumtak Ventures, and the European Innovation Council Fund. Total funding now stands at approximately €36 million.

Pind's assessment is that AI models and devices performing well in labs fail when confronted with reverberation, background noise, or unfamiliar physical environments.

Sound is a physical phenomenon, not just a signal processing problem. Treble's approach is based on physics-based acoustic modeling, simulating how sound propagates through space. It combines acoustic simulation, digital twins, and synthetic data generation, allowing developers to evaluate product performance before building physical prototypes.

This differs from traditional audio processing, which focuses on signal-level noise reduction, enhancement, and recognition. Treble instead computes sound propagation from spatial geometry, material properties, and source positions. Wall materials, furniture placement, and room dimensions all affect reflection and reverberation patterns. Inputting these parameters into the model generates acoustic data that closely mimics real spaces.

Amazon and Logitech are already using Treble's platform to test audio products. Wontak Kim, Amazon's senior audio research manager, says using virtual acoustic environments to evaluate audio quality and test scenarios that physical tests cannot replicate is a key part of developing Alexa. This highlights a practical need: many acoustic scenarios are difficult to reproduce or control in real environments.

For example, specific reverberation conditions, background noise at precise locations, and multiple simultaneous sound sources are costly to set up accurately in real rooms. Simulation environments allow these scenarios to be generated on demand and tested repeatedly.

Robots relying only on vision cannot hear what happens around corners. Paladin investor François Ruether notes that today's robots primarily use visual perception; they can detect what's ahead but cannot interpret sounds like someone falling in another room or a collision around a corner. This points to a capability gap in physical AI.

Vision is limited by line of sight—events behind corners, walls, or in another room are invisible to cameras, but sound travels. A complete perception system needs to process both seen and heard information for a fuller understanding of the surroundings. Treble's funding will focus on US market expansion and greater investment in physical AI, as voice and audio become key interfaces for intelligent machines.

Extending acoustic simulation from audio product testing to robot perception involves a transition. Testing speakers and headphones focuses on sound quality and noise reduction; robot hearing focuses on event recognition and spatial localization—such as direction, type, and whether a sound indicates an anomaly.

Whether Treble's existing physics-based acoustic modeling can support such tasks depends on whether its simulation data covers the sound scenarios robots will actually encounter. Amazon and Logitech use cases prove its value in consumer audio; the robotics direction is still early.