2.5 Years Later: A C++ Framework for Audio ML Research Prototyping - Let's break it down - Maxime Coutant - ADC 2025
---
After two and a half years as a R&D engineer in a public research laboratory, I'm happy (and stressed) to open source a big part of my work.
RT-Machine is a c++20 framework developed to help prototyping audio applications & libraries embedding neural networks. It targets linux/android/iOS to integrate ONNX/Tflite/executorch models and provide various features, like pre-processing and asynchronous processing.
We've been using it quite extensively to provide demonstration libraries to our partners.
In this talk I'd like to present you my work, but also discuss what it means to work as an engineer in a public research laboratory and what impact it had on the design of such a tool.
---
Audio enthusiast, hobbyist musician and software addict, I love to share, learn and meet new people !
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
Modular Synth in the Browser - Kilian Hertel - ADCx Copenhagen 2026
---
Synflow is a browser-native, visual node-graph workstation designed for the rapid prototyping and implementation of audio, MIDI, and control flows. Developed to fill the ""middle ground"" gap between low-level coding environments like the raw Web Audio API and complex, desktop-only legacy software, Synflow provides a high-speed playground for programming and DSP ideas. The platform prioritizes a zero-barrier entry, requiring no installation or account creation, and runs directly in any modern browser. Users can iterate quickly by adding and wiring independent nodes to hear results instantly, mimicking the tactile experience of a Eurorack modular system without the overhead of a traditional build setup. The technical architecture of Synflow consists of four distinct layers: the Web Audio API for the C++ render thread, XYFlow (React Flow) for the visual UI, a custom EventBus for cross-node messaging, and an Audio Graph Manager to keep the visual and audio planes in sync. With a library of 46 nodes, the system supports a wide range of functions including audio sources (oscillators, samples, noise), MIDI sequencing, logic operations, and audio transformations like filters and reverbs. Advanced users can even implement custom DSP via AudioWorklets or JS transforms $N rightarrow M$. Whether building generative music, live effect banks for instruments, or complex synthesizers like organs and bagpipes, Synflow offers a robust, open-source environment for both educational experimentation and professional-grade audio development.
---
Kilian Hertel
Kilian is a Product Manager for Software Development Services at Qvest Digital AG, where he currently focuses on leveraging AI to optimize development workflows. His career is a masterclass in bridging the gap between high-level strategy and deep technical execution, having spent nearly six years as the CEO of Tobax and serving as a Senior Developer at Tessitura GmbH.
Kilian’s philosophy centers on creating ""middle-ground"" tools that eliminate barriers to entry—a mission most visible in his project Synflow, a browser-native modular synthesizer that enables rapid audio prototyping without the need for installation or complex compile steps. With a background rooted in data modeling, project management at Deutsche Post AG, and full-stack development, he combines the reliability of industrial-scale IT with the creative agility of a builder who isn't afraid to ""wire it and hear it"" instantly.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
Using Strongly-Typed Units in Digital Audio Software - Roth Michaels - ADC 2025
---
API or math mistakes with units can cause problems ranging from a digital audio processing outputting silence to crashing your Mars rover—we’ll discuss real-life examples of both! There are efforts to solve these problems in the C++ standard library and open-source libraries that can be used today by audio developers.
The combination of user-defined types, conversion operators/constructors, and operator overloading in C++ give us the tools to use strong-types and avoid unit mistakes; std::chrono is a great example of this that everyone should be using. Unfortunately, when dealing with units beyond time many developers still use primitive types encoding units in variable names or comments.
This talk will provide an overview and update on progress of the proposed units library described in “P3045: Quantities and units library” which is currently going through standardization for C++29 and an update on progress. There will also be an introduction to mp-units, the open-source implementation of the proposed standard library features that you can use today.
To demonstrate how the proposed standard library can be extended to custom domains, extensions implementing various quantities and units used in digital audio / DSP that go beyond “physical” units. This will be code you can use today using mp-units and we are looking for industry experience and feedback!
We will close with a request for feedback and questions to see if these APIs will work for you.
---
Roth Michaels
Roth Michaels is a Principal Software Engineer at Native Instruments, an industry leader in real-time audio software for music production and broadcast/film post-production. In his current role he is involved with software architecture and bringing together three merged engineering organizations and legacy codebases: Brainworx, iZotope, and Native Instruments. He also supports the Audio Research team to help accelerate moving research to productization and developing fast prototyping tools for product teams. Before merging with Native Instruments, when he joined iZotope, Roth was the lead library designer of a new internal cross-platform ""Glass"", part of which is now available as open-source. More recently in his former role as Mix/Master Software Architect, Roth helped develop the reference implementation to move iZotope's products to subscription and led the team that launched the company’s first SaaS offering for music producers. Roth studied music composition at Brandeis University and continued his studies in the Dartmouth Digital Musics program. Roth began his career in software development writing software for his own compositions, and the works of other composers and artists, and teaching MaxMSP to composers and musicians; both private instruction and designing university courses. Before joining iZotope, he was working as a consultant for small startups working on mobile applications specializing in location services and Bluetooth.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
CTAG-TBD is an open-source embedded audio platform for low-latency real-time DSP, rapid algorithm prototyping, and deployment in Eurorack and standalone musical instruments. Developed within the Creative Technologies Arbeitsgruppe at HAW Kiel, the project builds on earlier ESP32-based audio work and provides a modular C++ architecture with dynamically loadable plugins and a simulator for hardware-independent development.
The platform explores the practical limits of commodity microcontrollers for deterministic audio processing. Implementations across ESP32, ESP32-S3, and ESP32-P4 address constraints including shared FPU resources, cache stalls during flash access, limited IRAM, PSRAM bandwidth, memory fragmentation, and cross-core interference. TBD’s real-time architecture isolates non-deterministic services such as the web UI, hardware control, and API handling from the audio path by assigning a dedicated core to DSP execution and using queues, atomics, IRAM-resident code, and custom arena allocation for robust runtime plugin management.
Recent ESP32-P4-based prototypes demonstrate substantially improved computational and memory performance, dual-core floating-point execution, SD-card-hosted UI assets, and stable sub-millisecond audio latency with dynamically loadable DSP modules. Combined with a CMake-based workflow, JSON-defined interfaces, an integrated DSP library, and VCV Rack support, CTAG-TBD provides a practical research and development platform for embedded audio, open musical instrument design, and musician-centered interaction.
---
Robert Manzke
Robert is Professor of Computer Science at HAW Kiel, where he focuses on ubiquitous computing, creative technologies, and music technology. With a background in electrical engineering, DSP, and a Ph.D. in computational imaging sciences, his work bridges signal processing, algorithm development, and interactive embedded systems. His lifelong interest in synthesis and electronic music led him to found CTAG, the Creative Technologies Arbeitsgruppe at HAW Kiel, in 2013 as a university R&D group for student-driven projects in the domain of technology for creativity.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
Doodads & Doohickeys - A Guide to Theory-Crafting & Modelling a Fictional Acoustic Instrument - Syl Morrison - ADC 2025
---
Physical modelling is a theory-dense and complex area of plugin design, which tries its utmost to remain faithful to the real-world mechanics of the modelled instrument. But what if that instrument doesn't exist?
This talk aims to explore this concept, prioritising a ""this sounds fun"" approach, rather than a ""this is accurate and actually real maths"" approach, in a hand-wavy, conversational and engaging way. Its purpose is less to give practical modelling advice, and more to encourage experimentation, and to celebrate the art of exploring bizarre-fever-dream plugin ideas.
---
A self-taught audio software developer fascinated by eldritch, abstract and unconventional effects, based in Dublin, Ireland.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
Reshaping the Voice - Real-Time Neural Audio Effect - Anders Bargum - ADCx Copenhagen 2026
---
In this presentation, Anders Bargum introduces Reshaping the Voice, a real-time neural audio effect designed for voice conversion and speech modification. Voice conversion involves altering a speaker's vocal identity while keeping the underlying words and linguistic content intact. While state-of-the-art voice generation models achieve highly realistic results, they are heavily constrained by high parameter counts, non-causal processing, and high latency, making them unusable for live applications such as digital audio workstations (DAWs) or interactive virtual environments.
To overcome these limitations, the project combines speech representation learning with musical timbre transfer. By utilizing feature disentanglement, the model disassembles input audio into independent streams representing vocal identity (timbre), linguistics (phonetic units), loudness, and pitch. These features can then be individually manipulated, enabling real-time identity swapping, pitch shifting, and even the generation of multi-part vocal choirs from a single monophonic input.
The technical implementation leverages a streamable pipeline trained in PyTorch and exported via TorchScript to run efficiently inside a C++ framework hosted by JUCE. By implementing causal convolutions and state caching rather than relying on future-sample lookaheads, the system achieves incredibly low latency. Anders demonstrates practical deployment of this technology, showcasing a fully functioning VST plugin and an integration within the Unity game engine for designing real-time voices for 3D avatars in virtual reality.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
It’s Just a Phase - Exploring Synthesis With the Phase Vocoder - Cameron Thomas - ADC 2025
---
Can we do time-stretching…in real-time? What if instead of just stretching time, we could freeze it, scrub through it, and morph between different moments in a sound?
We will explore this question, and more, by trying to build out some creative ideas for synthesis armed with just a few samples and the trusty FFT.
We will cover the basics of time-stretching audio using the FFT with phase-vocoding, and in doing so try to build an intuition of how phases in a Fourier Transform affect the sound with some beautiful interactive plots.
We'll then try to build out a full synth using some of these (usually offline) time-stretching techniques - a waveform player where you can scrub over the sound at any speed, use a playhead to pause, freezing the sound, then randomly move around. We will introduce some pitch shifting, and show that phase-vocoding offers a texturally rich and organic alternative to granular synthesis - no grain boundaries or windowing artifacts.
Through the whole talk I'll try to answer questions like:
- What the heck even ARE phases?
- How do we handle phase coherence differently for onsets versus harmonic sounds?
- How do we make processed noise sound natural instead of artificial?
- How can we use these algorithms in real-time synthesis?
Throughout, we'll dig into practical implementation details - you'll come away with some tools for using FFTs in ways you might not have expected, and techniques for making these algorithms work in real-time.
Target Audience: Audio developers with basic FFT knowledge interested in creative synthesis applications and practical DSP implementation.
---
Cameron Thomas
Hi, I'm Cam.
I run Minuit, a small indie plugin company where I'm currently releasing (or just released, hopefully) Solstice - an experimental synthesiser based on the phase-vocoding techniques I'll be talking about today.
Previously, I spent six years as a Staff Software Engineer at Vochlea Music, developing the Dubler voice-to-MIDI applications. These days I've gone indie to explore more experimental sound design ideas, whilst also taking on contract plugin development.
I'm particularly interested in making complex DSP feel intuitive and playable. As someone who makes experimental and ambient music in my spare time, I enjoy finding new ways to process sound that go beyond traditional synthesis methods - which is exactly what led me down the phase-vocoder rabbit hole, after playing around with neural synthesis back in 2018/19 and finding it frustrating to work with.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
Cross-Platform Framework for Textural Granular Processing - Aman Jagwani & Victor Lazzarini - ADCx Copenhagen 2026
---
This talk presents a cross-platform C++ framework for real-time textural time-domain and spectral granular processing called Granular Recipes.
Granular synthesis is a well-explored area of audio DSP that exposes a large number of parameters to users, such as density, pitch, grain size, position, and jitter, resulting in a wide range of sonic possibilities from dense clouds to rhythmic stutters. While this breadth is what makes granular synthesis so powerful, the abundance of parameters and possible sonic outcomes can also lead to a loss of focus or an overwhelming decision-making process, especially during live performance.
Our framework proposes a different approach to granular program development and performance. Instead of exposing raw parameters, we introduce a system of “recipes” or behavioral structures that developers write to define how granular parameters relate to each other and evolve over time. A recipe functions like a developer-level preset that provides a single sonic identity to the resulting granular synthesis, such as pitch-shifting micro-loops, sparse textural bursts, dense clouds, or tempo-synced stutters. This approach allows the developer to constrain or choreograph the parameter space, embedding a sense of compositional intent directly into the programming process.
Once a recipe is active, the performer controls the evolution of texture through just two macros: energy and intensity. Modulating these macros animates the sound while remaining within the sonic signature defined by the recipe. The use of only two macros also makes the resulting instrument particularly conducive to gestural and sensor-based control, since only two mappings, and therefore only two sensors, are required to produce comprehensive transformations of the sound.
In the talk, we present the framework’s layered architecture, consisting of a platform-agnostic core DSP engine, the recipe API, and thin platform bindings that allow the same C++ recipe code to run on desktop, microcontroller, and FPGA targets. On FPGAs, the grain engine runs as custom IP modules in the programmable logic, enabling ultra-low latency and high parallelism, while the recipe system runs on the ARM processing system. By leveraging the throughput and parallelism offered by the FPGA, we have also implemented spectral granular processing, which can interact with the same recipe API, extending the system into the frequency domain within this embedded environment.
---
Aman Jagwani
Aman Jagwani, from Mumbai, India, is an audio programmer, musician, and researcher. He completed his Bachelor's degree in Electronic Production and Design and Drum-set Performance at Berklee College of Music, Boston, and later earned a Master of Science in Sound and Music Computing from Maynooth University.
Currently, Aman is undertaking his Ph.D. at Maynooth University with supervisor, Prof. Victor Lazzarini, for which he received the Hume Fellow Scholarship Award. His research centres on audio DSP, embedded systems, FPGAs and interactive musical experience design.
Victor Lazzarini
Victor Lazzarini is a Professor of Music at Maynooth University. A BMus graduate of the Universidade Estadual de Campinas (UNICAMP) in Brazil, he received his AMusD from the University of Nottingham, UK. His academic interests include musical signal processing and sound synthesis; computer music languages; electroacoustic and instrumental composition, and improvisation. He has authored over one hundred fifty peer-reviewed publications in his various specialist research areas. He is the author of Aulib and Aurora, two object-oriented libraries for audio signal processing, and is one of the project leaders for the Csound sound and music programming system.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
In this talk from ADCx Copenhagen 2026, electronic musician, composer, and audio programmer Cristián Vogel introduces Incline, a native desktop instrument developed at his studio, Never Engine Labs. Moving beyond traditional time-ordered slicing and blind segmentation found in many granular synthesis tools, Incline treats the act of segmentation as the primary musical decision, putting compositional control back into the hands of the creator.
The core of Incline lies in its multi-dimensional analysis stage, where spectral data functions as the absolute source of truth. By processing fixed stereo files through a principal component analysis (PCA) algorithm, the software maps audio fragments onto a non-linear, navigable "topersonic" terrain based on spectral similarity rather than a chronological timeline. Vogel demonstrates how composers can perform these sounds by navigating the 2D surface manually, via MIDI, or through a built-in contour sequencer, offering three distinctive runtime behaviors—one-shot, one-loop, and multi-layer cloud modes—that allow similar spectral clusters to be recontextualized musically.
Vogel also shares the underlying development story, highlighting how modern frameworks like Rust, Fun DSP, Tauri, and the FluCoMa library from the University of Huddersfield made the software possible. Reflecting on the evolving landscape of software production, he explores the rising significance of AI-assisted development ("vibe coding"), which enables contemporary artist-engineers to bridge the gap between creative ambition and complex digital signal processing to bring entirely new instruments to life.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
Building an Optimized DSP Framework in Modern C++ - Scott Carver - ADC 2025
---
The introduction of C++23 and upcoming C++26 features have significantly expanded the potential for building expressive, manageable embedded domain-specific languages (EDSLs) that compile efficiently into optimized lower-level representations. Features like enhanced constexpr and consteval execution enable a single, descriptive and clear representation of audio DSP algorithms to generate multiple optimized versions tailored to specific execution contexts and runtime constraints.
Inspired by successful libraries such as Halide and Eigen, we will look at a prototype C++ library developed to explore clean separation between algorithm description, data structures, and execution strategies. It generalizes inner-loop optimizations vectorized across two dimensions, enabling consistent application of advanced optimization techniques across diverse DSP algorithms. The library provides us a collection of design patterns that we can use to formulate how a contemporary C++23/26 audio graph can be built.
This talk will explore practical techniques and simple examples for implementing these design patterns in modern C++, demonstrating how they collectively form a robust framework for defining and executing efficient DSP algorithms:
- Zero-cost Compile-time / Runtime Variant Values: Approaches to building scalar compile-time-constant and runtime-variant values, as well as more complex aggregate / property tree data structures, without incurring memory or performance penalties.
- Parametric Data Structures: Designing simple, readable data structures that can be optimized cache friendly access and vectorization.
- Expression Templates: Using C++'s template metaprogramming to defer computation, enabling optimized evaluation and graph transformations.
- Loop and Data Optimizations: Techniques for loop-invariant value hoisting, re-blocking, and linearization to improve memory and execution efficiency.
- Finite Difference Optimization: Automated optimizations for polynomials, trigonometric functions, and interpolation routines to minimize runtime computation.
- Unified Parallelism Models: Integrating voice-parallel and time-parallel computations within a single descriptive graph representation.
- Compile-time Graph Transformations: Leveraging constexpr to construct and optimize directed graphs representing DSP signal flows, including handling feedback loops and delays.
- Benchmarking Execution Strategies: Describing, benchmarking, and comparing graph execution strategies (time-parallel, voice-parallel, block-based), enabling parametric, empirical optimization decisions.
- Empirical Validation: Investigating whether and how these techniques improve DSP performance in realistic scenarios, with comparisons against established commercial and OS audio engines.
By applying these strategies, we demonstrate how a single algorithmic representation can flexibly handle multiple DSP use-cases—from polynomial interpolation and envelope generation to buffer read/write operations—significantly reducing redundant code and enabling focused optimization.
Furthermore, separating algorithm logic from execution strategies allows parameterizing benchmarks and optimizations based on CPU architecture and usage patterns, preserving algorithm clarity without embedding machine-specific details.
These strategies will be presented individually so they can be applied immediately in existing audio code-bases, and demonstrated within a unified framework context to highlight potential improvements in code clarity, performance, maintainability, testing, and benchmarking. Topics will be accessible for moderate-to-advanced C++ programmers, and anyone interested in exploring cutting-edge practices in modern audio DSP programming. Attendees will gain actionable insights, directly applicable techniques, and inspiration for advancing audio DSP code performance and clarity with modern C++.
---
ADC is an annual event celebrating all audio development technologies, from music applications and game audio to audio processing and embedded systems. ADC’s mission is to help attendees acquire and develop new audio development skills, and build a network that will support their audio developer career.
Annual ADC Conference - https://audio.dev/
https://www.linkedin.com/company/audiodevcon
