Treble Technologies sells cloud software and synthetic datasets that help device makers and AI developers test how microphones, speakers and voice models behave outside a quiet lab. The Icelandic company has raised an $18 million Series A-2 led by Paladin Capital Group, with existing investors KOMPAS VC, Frumtak Ventures and the European Innovation Council Fund also participating.
An engineer can import a device design, place microphones and sound sources, and define a room or moving environment. Treble’s engine then models how sound travels, varies the conditions and produces labeled audio that can feed training or evaluation. The same setup lets teams compare hardware and algorithms before building more prototypes or organizing recording campaigns.
The distinction matters because a voice system that works in controlled conditions can stumble on reverberation, background noise or a different microphone position. Recording every room, vehicle and edge case is slow; simulation makes those conditions repeatable. Treble says its platform can model rooms, materials, devices, speakers and ambient sounds, then run many variants in parallel.
Treble sells access through a web application, a Python-based development kit and prepared acoustic datasets. Its web application is sold by subscription, while enterprise, development-kit and dataset access is available through custom plans. Amazon and Logitech are named customers: Amazon uses virtual acoustic environments in Alexa product and model development, while Logitech uses the software to test conditions that are difficult to reproduce physically.
The round follows $12 million raised in 2024 and takes disclosed funding above $40 million, according to TechCrunch. Treble said the new money will support United States growth and expand its work with robotics, automotive and drone developers. Those markets add a harder variable: the listening device itself may be moving while it tries to interpret speech or a warning sound.
The push isn’t only about generating training data. Treble also evaluates models against repeatable acoustic scenarios; its partnership with Hugging Face produced a 2026 benchmark for speech-recognition systems under far-field, noisy and reverberant conditions. That gives model makers a common test for the failures Treble’s simulations are designed to expose.