# Atlantic Audio Case Studies

> Atlantic Audio case studies: Playbeat 4 for Audiomodern, the DiscoDSP synthesizer interfaces, and a confidential neural vocal restoration sprint in PyTorch.

- Page: https://atlanticaudio.dev/case-studies
- Agency overview: https://atlanticaudio.dev/index.md
- More shipped work: https://atlanticaudio.dev/#portfolio
- Contact: info@atlanticaudio.dev

Each case study here records one engagement: what the client needed, what we owned, how we built it, and what shipped. Named projects appear with the client's permission. Our engagements are confidential by default, so many stay unlisted, and one of the three below is published anonymized.

## The written-up engagements

- [Playbeat 4 (client: Audiomodern)](https://atlanticaudio.dev/case-study/playbeat-4.md): Audiomodern's AI-powered drum designer, released on 5 March 2025. We built the full application and its sequence randomization engine in JUCE and C++, from the sequencer core to every desktop, iPadOS and Logic Pro MIDI FX target. Stack: JUCE, C++, VST3, AU, AAX, AUv3, iPadOS. Outcome: improved performance and stability across complex sessions, reduced CPU spikes, and crashes eliminated in high-load scenarios. HTML page: https://atlanticaudio.dev/case-study/playbeat-4
- [OB-Xd, Discovery Pro, Corona (client: DiscoDSP)](https://atlanticaudio.dev/case-study/discodsp.md): Three DiscoDSP synthesizers that ship as VST, VST3, AU and AAX. We rewrote and maintained their interfaces in C++ through 2021 and 2022, covering panel layout, custom controls, visual feedback and GUI scaling, and left the synthesis untouched. Stack: C++, UI/UX, VST3, AU, AAX. Outcome: drastically improved visual feedback, smoother interactions, and a renewed lifespan for the plugins. HTML page: https://atlanticaudio.dev/case-study/discodsp
- [Neural Vocal Restoration (client: confidential)](https://atlanticaudio.dev/case-study/neural-vocal-restoration.md): A 2026 fixed-fee feasibility sprint in Python and PyTorch for an independent artist, aiming for raw close-microphone vocal texture in AI-assisted renders while keeping the voice identity and the performance intact. Stack: Python, PyTorch, ML. Outcome: clearly audible gains in vocal texture and naturalness over the stock reconstruction route, with the voice identity and the performance intact. HTML page: https://atlanticaudio.dev/case-study/neural-vocal-restoration

## The rest of the shipped work

Selected Works on the home page carries the remaining portfolio: WaveBalance for CodWaves, LatencyFix, the Djaminn app and an embedded neural network built for a confidential client. Those entries are short summaries; the three pages above are the long form.

- Selected Works: https://atlanticaudio.dev/#portfolio
- Agency overview: https://atlanticaudio.dev/index.md
- Start a project: https://atlanticaudio.dev/#contact
- Book a call: https://calendar.app.google/QvJwizyxHyndJMea7