Icelandic startup Treble has raised $18 million to expand a simulation platform designed to help AI companies test how speech and audio systems perform outside controlled laboratory conditions.
TechCrunch reported that the $18 million extension to Treble’s Series A was led by Paladin Capital Group, with KOMPAS VC, Frumtak Ventures, the European Innovation Council and Omega participating. The new funding brings the company’s total capital raised to more than $40 million.
Treble was founded in 2020 by acoustic engineers Finnur Pind and Jesper Pedersen and counts companies including Amazon and Logitech among its customers.
Voice AI has a real-world testing problem
Voice models can perform well on clean recordings but struggle when speakers are far from microphones or surrounded by noise and reverberation.
Treble is trying to simulate those difficult environments rather than forcing developers to gather every possible acoustic condition through physical recordings.
Pind told TechCrunch that audio AI is really a data challenge and argued that physics-based simulation could provide an alternative to relying only on recordings or audio scraped from the internet.
The company provides synthetic-data generation for applications including speech enhancement, noise suppression and model training. It also lets companies evaluate voice systems under different acoustic conditions before deploying them.
Earlier this year, Treble worked with Hugging Face on the FFASR leaderboard, which tests speech-recognition systems under far-field conditions including noisy rooms, reverberation and difficult signal-to-noise ratios.
The benchmark uses the same held-out data and text normalization for participating models so their performance can be compared across acoustic conditions.
Simulation is becoming infrastructure for AI testing
Treble is also moving beyond speech-model evaluation.
TechCrunch reported that the company is working with headphone, speaker and smart-device companies on virtual prototyping and is expanding into robotics, automotive systems and drones, where sound can become another source of environmental information.
Paladin Capital Group vice president Francois Ruether shared that the firm expects acoustic simulation to become more valuable as products increasingly depend on understanding sound across voice AI, wearables, robotics and physical AI.
That makes the funding more interesting than another $18 million AI round.
Much of the AI infrastructure market has focused on generating more training data and supplying more compute. Treble is targeting what happens afterward: whether systems trained in controlled environments actually survive contact with messy physical environments.
For builders, the larger signal is that AI testing itself is becoming an infrastructure category. As voice interfaces move into more products, realistic simulation may become as important as the model being tested.
