Billions of dollars are pouring into this field, from automated customer service and sales call transcription to the proliferation of meeting recorders, and even AI smart glasses that use sound as the primary interface. At the same time, major AI labs are accelerating model iterations, while hardware manufacturers are striving to deliver the best audio interaction experiences for users. However, all these new technologies require rigorous testing and feedback loops to continuously optimize them. The Icelandic startup Treble is trying to become the core infrastructure behind sound technology for large model vendors, robotics companies, and consumer hardware manufacturers by building a dedicated simulation platform.

The company was founded in 2020 by acoustic engineers Finnur Pind and Jesper Pedersen. Recently, Treble successfully completed an $18 million Series A extension round. This round was led by Paladin Capital Group, with participation from existing investors KOMPAS VC, Frumtak Ventures, EIC, and Omega ehf. Adding the $12 million investment it received in 2024, the company's total funding so far has exceeded $40 million. Its client list now includes industry giants such as Amazon and Logitech.

In terms of specific product areas, Treble's business mainly revolves around simulation and data generation. For voice AI companies, it launched a synthetic data generation platform, which can be widely applied to voice enhancement, noise suppression, and model training. At the same time, the platform can also test and evaluate the performance of voice AI models in various complex environments, providing detailed feedback to laboratories. Notably, earlier this year, Treble also collaborated with Hugging Face to launch an evaluation benchmark for speech recognition models under different real-world scenarios.

Co-founder Finnur Pind said that the core challenge of voice AI is fundamentally a data issue, and this is also the key opportunity for the next generation of models and hardware. For a long time, almost all voice-related AI has been trained using recordings and data scraped from the internet, but Treble believes that high-precision physical simulation can become a new alternative path for creating voice data.

Aside from software-level data and evaluation, the startup also focuses heavily on hardware-level sound design and testing. For example, Treble works closely with headphone and speaker manufacturers, helping them accurately assess the actual sound quality of products through virtual prototyping, or testing the command recognition effect of smart speakers in different placement positions. Recently, Treble's business scope has further expanded into the simulation testing of smart glasses and various AI wearable devices.

Speaking about future technological vision, Finnur Pind expressed high expectations for the next generation of wearable devices. He believes that devices like earphones and smart glasses could one day achieve functions similar to "super hearing," helping users hear more clearly in complex acoustic environments. For example, amplifying conversations within two meters in a noisy restaurant, or automatically silencing surrounding noise during a workshop.

Looking ahead, Treble plans to shift its focus further toward the field of embodied artificial intelligence, targeting services for robots, cars, and drones. Through testing and simulation, it aims to help these physical hardware better implement various sound-based functions.

Franois Ruether, Vice President of Paladin Capital Group, said when talking about the investment logic, that Treble's platform, which can simulate different models and devices, has extremely high uniqueness in the industry. As the company expands into more fields, the importance of the platform will also increase. In his view, as more and more products begin to rely on understanding sound, this infrastructure will increasingly highlight its value in the fields of voice AI, wearables, robotics, and embodied AI. Throughout this process, customers can always retain ownership of their models, products, and development workflows, while benefiting from the shared empowerment provided by the acoustics infrastructure layer built natively on simulation.