Treble: 2023 in review and 2024 outlook

There are no two ways about it, 2023 was a wild ride for us at Treble: the launch of our platform, the on-boarding of more than 100 organizations as paying customers (including many of the biggest tech companies and engineering firms in the world) and growing the Treble team by more than 50%. Winning the startup competition at one of the biggest tech conferences in the world: the International Congress on Acoustics, Speech, and Signal Processing (ICASSP 2023) and being shortlisted by the Financial Times as Tech Champions of 2023.

We founded Treble back in 2020 with a clear vision, as summarized in the slogan we have used from day one: Enabling a Better Sounding World. More specifically, to build next generation software tools that will enable engineers and designers across different sectors to make the world a better sounding place. Because, as research has demonstrated repeatedly, better sound and less noise results in improved health, well-being, productivity and improves people’s ability to communicate and connect with one another. In fact, noise and poor acoustics are, according to the WHO, the second most harmful environmental factor impacting our health, so there is a lot to be gained from a better sounding world. We at Treble believe that accurate, efficient, and easy-to-use sound simulation tools can be a catalyst for realizing better sound experiences, and considering the massive traction Treble has seen in just 9 months from launch, it seems like the industry shares this view with us.

The R&D efforts behind Treble can be traced all the way back to 2013, the year Finnur Pind, wrote his first wave/geometrical hybrid acoustic simulation algorithm, but things properly kicked into gear in 2020, and finally 2023 was the year where version 1.0 of the Treble sound simulation platform was brought to the market. On March 1st we launched the graphic user interface version (the Treble Acoustic Simulation Suite) and on October 17th we launched the programmatic interface version (the Treble Software Development Kit).

The Treble Acoustic Simulation Suite

The Simulation Suite targets, at least here in the beginning, room acoustics design and electro-acoustic system design. What we are bringing to the table with this product is three things:

• Better simulation accuracy. The Simulation Suite is powered by Treble’s proprietary simulation engine, which combines an ultra-efficient wave solver for low-mid frequency modeling and a state-of-the-art geometrical acoustics solver for high frequencies. The wave solver has been benchmarked to be at least 100x faster than existing desktop-based solutions. What this means is that wave modeling can now finally easily be used in practice, even in large scale problems, which greatly improves simulation accuracy. This, in turn, reduces risk during the design process and makes it easier to realize high quality designs. It has been thrilling to carry out numerous benchmark studies over the year where we see that Treble consistently outperforms existing solutions in terms of accuracy, see here: https://docs.treble.tech/validation/
• Better workflow. Anyone who has worked with designing room acoustics and electro-acoustic systems knows that there is so much wasted time and effort on mundane manual work, trial & error, and 3D model clean-up. Our mission is to remove all this wasted effort so that users can focus on the things that matter: the actual design work. We have made meaningful progress, but there are also major workflow breakthroughs in the pipeline (more on that below). The Simulation Suite is the world’s first browser-based and cloud-based sound simulation product. This means that simulations can be launched from any device, and that virtually unlimited simulations can be run simultaneously. We have built a live-link plug-in into SketchUp, which removes the need for 3D CAD file exporting/importing and makes iterative design changes a breeze. We have built a beautiful and easy-to-use user interface and we have introduced cloud-based collaboration features, something which has not existed before in the audio/acoustics world.
• Better auralizations. The beauty of working with acoustics is that you can use your ears to gauge the quality of your design: the concept of auralizations. With the Simulation Suite it has become easier than ever to set up virtual listening experiences where users can immerse themselves in the 3D models, move around the model and compare design options on the fly. The latest addition is the ability to create shareable auralizations that users can send through the web, which greatly facilitates communicating acoustics designs with clients and other stakeholders.

Since the v1.0 launch on March 1st, countless features and improvements have been released, with updates being pushed to users nearly every week. These include:

• Surface receivers – heatmap results to explore the spatial variation of sound in space
• Loudspeaker modeling – introducing the capability to import loudspeaker characteristics via the CLF file format and including a large database of loudspeakers
• Improved solver performance – we sped the wave solver up by 50% and the geometrical acoustics solver by orders of magnitude
• Collaborative features – allowing Google-Docs style collaboration within organizations, e.g., shared material and loudspeaker databases and shared projects
• Automatic geometry simplification – now users can import “imperfect” 3D models (e.g., non-watertight) and run simulations on them
• Shareable auralizations – with a click of a button send a virtual listening experience of your designs to stakeholders across the globe
• Open office parameters – predict acoustic parameters for open office environments according to ISO 3382
• Updated anechoic sound library + ability to import own sounds into the auralizer
• Results view overhaul – making it easier to explore and compare the results of simulations
• Autostop – users no longer need to set the expected IR length of simulations
• Countless UI/UX improvements

It has been exciting to witness the reaction of the industry to this product. The industry is calling for a modernization of the workflow when doing acoustics design. Moreover, many users are enthusiastic about finally having the ability to do wave modeling efficiently and we expect wave modeling to steadily become the industry standard simulation approach in room/electro-acoustic simulations. We at Treble are so lucky to be bombarded with passionate feedback from the community, both praise but also lots of great suggestions for improvements and new features. We take the feedback very seriously and it has a significant impact on our roadmap, so please keep the great feedback coming! Here are some examples of user testimonials from the year:

“Due to the speed, user-friendliness, and accuracy, we have been able to provide more simulation services in three months than we have with other software packages in probably 6 to 8 months. Treble has increased the productivity of our room acoustic engineering team in an enormous way.”

Bradley Alexander, Acoustic Engineer, Bollinger + Grohman

“Before Treble, wave-based acoustic simulation was too slow and cumbersome to use in anything but very small rooms, while geometric acoustics platforms couldn't produce accurate data at low frequencies and could take days to calculate a complex set of auralization-quality impulse responses. With Treble, wave-based and geometric acoustics are seamlessly blended behind a clean, modern GUI and the blazing fast cloud processing does in an hour what used to take me a week. Treble's speed and ease of use let me use acoustical simulation on a wider range of projects than any other tool currently available.”

Matthew Azevedo, Sr. Scientist, Acentech

“The Treble Acoustic Simulation Suite is a groundbreaking piece of software, allowing us to work much faster, quickly hear the results via auralization and be much more confident about our acoustic designs. Being well integrated with SketchUp makes it very easy and fast to create/modify acoustic models. The intuitive workflow and fast cloud computation allows to go from importing a geometry to having results in only minutes.“

Jannik Enders, Acoustic Engineer, Akustik Bureau Dresden Ingenieurgesellschaft

“In Treble it is extremely easy to create, manipulate, and share audio and predictive demonstrations of room acoustics with unheard-of speeds. My clients love the ability for us to collectively manipulate their designs in real-time and then see/hear the results in mere minutes. We have had a very positive experience showcasing this software as part of our acoustical consulting practice.“

Ethan Bourdeau, Owner and Principal Acoustician, Bourdeau Acoustical Design

“In addition to the truly impressive auralizations and the ability to predict the frequency response at the mix position, Treble offers us incredible speed and efficiency in the creation and analysis of the data. With Treble, we have managed to take our acoustic studies to a level that really sets us apart from the others in terms of quality, flexibility, and communication of results.”

Philip Puzalowski, Acoustic Engineer, Audics

The Treble Software Development Kit

The Treble SDK was launched in Q4 2023. A truly first-of-its-kind product, it offers a Python-based programmatic interface to Treble’s cloud-based sound simulation engine. We have been overwhelmed with interest in this product, as the use cases for such a product are countless and the flexibility of the programmatic interface opens so many avenues to do things that simply were impossible to do before.

Synthetic audio data generation

The audio industry has drastically increased its usage of machine learning based methods in recent years and I fully expect this trend to continue. This applies, e.g., to various real-time spatial audio signal processing applications such as speech enhancement, dereverberation, room compensation and source localization. The industry is also moving towards data-driven methods for real-time audio simulations and rendering immersive experiences, e.g., in AR/VR.

As with all machine learning applications, the performance of the ML system will depend on the quality of the data it is trained on. When it comes to audio data collection/generation, the industry has up until now primarily relied on either doing real-world data collection through measurements or synthesizing spatial audio data using simple simulation methods. Real-world data is generally of high quality, but data collection is expensive and capturing edge cases is challenging. Synthetic data generation is cost efficient, but research has shown that the synthetic audio data that state-of-the-art methods can produce is not of high enough quality and the resulting ML algorithms do therefore often not generalize very well to real world scenarios.

With the Treble SDK, it becomes feasible to generate large volumes of spatial audio data in realistic and complex acoustic environments using Treble’s high precision simulation methods at scale. We have seen robust evidence of this high-quality synthetic data yielding greatly improved ML performance over simplistic synthetic data generation methods (one example here on blind room estimation, where errors were reduced by 50%: https://appfa2023.silsystem.solutions/atti/000653.pdf), which in turn implies that synthetic data generation can augment or replace real-world data collection. Treble is working with many of the big tech giants and large names in the audio industry on this exciting and emerging use case.

Virtual prototyping & advanced simulations

Another exciting use case is that of virtual prototyping of audio devices in different acoustic conditions, e.g., assessing microphone array performance or downstream tasks such as speech recognition. Virtual prototyping implies replacing/augmenting physical testing and physical prototypes with virtual testing using simulations. Virtual testing is much more cost-efficient, enables faster design cycles and more design iterations – but it is only a viable approach if the simulation results are reliable. With the Treble SDK, it becomes possible to run high-precision full-bandwidth wave simulations in large and complex scenarios within practical runtimes and output ultra-high order spatial IR output. This enables reliable virtual prototyping of various audio devices, again something that many of the big tech giants and large names in the audio industry have started to use the Treble SDK for.

Automated simulations, parametric studies, and 3rd party tool integration

The programmatic interface that the SDK offers allows users to use scripting to set up simulations in an automated fashion. This means users can in a systematic way create and run simulations and analyze results, e.g., doing parametric studies while varying some simulation parameters (e.g., sound sources, receivers, materials, geometry, etc.) and even performing automated optimizations, where the ideal solution to a given problem is programmatically discovered.

The programmatic interface also allows users to plug the Treble simulation engine into 3rd party solutions and/or their own custom R&D workflows. An example is that of integrating the SDK into Rhino/Grasshopper (see demo here: https://youtu.be/bjh3NLYT7oo?si=UcHJuA4fjPP5oDZY&t=2801), where it becomes possible directly from Rhino to run simulations programmatically and get back results. Rhino/Grasshopper is just one example, the SDK can be integrated into gaming engines such as Unity and Unreal and other CAD modeling tools beyond Rhino – anything that speaks Python.

Conferences and trade shows in 2023

It was a blast for us Treblemakers meeting colleagues and partners across the globe at different conferences and trade shows, where we had booths and/or scientific presentations:

• Ecophon International Acoustics Symposium in Halmstad, Sweden
• Acoustical Society of America 184th Meeting in Chicago, US
• AES (Audio Engineering Society) Europe in Espoo, Finland
• International Conference on Acoustics, Speech and Signal Processing in Rhodes Island, Greece
• AES International Conference on Spatial and Immersive Audio in Huddersfield, UK
• International Conference on Immersive and 3D Audio (I3DA) in Bologna, Italy
• Forum Acusticum in Torino, Italy
• Audio Engineering Society Main Event in New York, US
• Tonmeistertagung in Duesseldorf, Germany
• Slush in Helsinki, Finland

Growing the Treble team

The Treble team grew by more than 50% in 2023, from just under 20 people to around 35 people. We added experts in machine learning, acoustic simulation science, geometry processing, pro-audio, cloud infrastructure, high performance computing and product design, as well as sales, marketing, and finances.

We plan to grow our team further and look forward to collaborating with our existing and new customers and partners! We will participate in many conferences and tradeshows, e.g., BNAM, InfoComm, Internoise, ICASSP, BIM World and AVAR. Be sure to reach out to us if you are interested in giving Treble a spin or if you have any questions or comments.

We at Treble ended the year by making a sizeable donation to humanitarian aid efforts by Unicef in the Gaza region, and we hope that 2024 brings an end to the conflict there and elsewhere in the world.