Case study Audio · Pro tools · Designed 2021 to 2022 · Read 2026

ELIAS 4: designing a middleware that composers can actually use.

End-to-end product design for a node-based adaptive audio engine: the strategy against Wwise and FMOD, the Studio's architecture, and the bets behind every screen.

RoleSenior Product Designer, contract, end-to-end
CompanyElias Software
TimelineApr 2021 to Mar 2022 · retrospective 2026
PlatformDesktop authoring + C/C++ runtime
No soundbanksIteration drops from minutes to seconds
One canvasSignal, logic, parameters and DSP in one graph
Live remotingEdits land in a running game instantly
PythonStudios script their own tools
In 30 seconds
  • Four workflow bets against Wwise and FMOD: no soundbanks, one node canvas, live remoting, Python designtime. None of them is an audio feature.
  • One mental model everywhere: four content types, a fractal graph, DAW conventions spent exactly where familiarity buys the most.
  • Outcomes: the design work anchored a successful funding round and the design function scaled from one designer to five.
ELIAS Studio patch editor with audio, logic and DSP nodes on one canvas
FIG. 01 The patch editor: signal flows left to right, typed color-coded ports make legal connections visually predictable, and playback never leaves the canvas.
Open the clickable Studio reconstruction
01 · Context

A category owned by two incumbents, and their shared weakness.

Role note. I designed ELIAS 4 end to end from 2021 to 2022: IA, core flows, the node editor, and the design system. This study pairs that work with a retrospective read of the shipped product as of 2026. The category is dominated by Wwise and FMOD, both powerful, both carrying decades of interface complexity: soundbank pipelines, rebuild cycles, and learning curves that skew the tools toward audio programmers instead of the composers and sound designers who author the content. That defined the brief: the entire competitive strategy rests on interface and workflow quality.

Personas reflect the two authoring roles the product was built for, drawn from studio partners and community feedback during development.

The core UX thesis

The incumbents' weakness isn't capability, it's friction. Every major feature decision attacks a specific incumbent friction point.

02 · Four bets

None of them is an audio feature.

No soundbanks

Kills the rebuild cycle that breaks flow on every content change; "fast workflow without soundbanks" is the headline, not a footnote.

Visual scripting

Signal, logic, parameters and DSP on one sample-perfect canvas that non-programmers can read and author.

Live remoting

Connect to a running game session and mix, swap, and tweak with instant in-game updates.

Python designtime

Studios script their own tools against live project data, so the tool becomes a platform.

Traditional middleware Edit Rebuild banks Export Build and launch Listen minutes per change ELIAS 4 Edit Hear it in the gamelive session, no banks, no rebuild seconds every product layer shortens this same loop
FIG. 02 The entire differentiation, drawn: traditional middleware makes you rebuild to listen; ELIAS closes edit-to-hear to seconds.
03 · The Studio

One mental model, four zones, fractal depth.

PATCH_A · PATCH_B · ARR_SIMPLEdocument tabs: graphs and timelines are peers Project browser PatchesMusicMixersAssets the whole mental modelis four buckets Tabbed canvas node graphs, arrangements and the profiler share one surface subgraphs open on the same canvas, one level deeper Inspector selection-driven parameters, transitions the canvas stays clean Transport and status: "can I hear it right now?" is always one click, the affordance behind the fast-iteration promise A LAYOUT ANY DAW OR IDE USER RECOGNIZES INSTANTLY, WHICH IS THE POINT
FIG. 03 The four-zone architecture. Familiarity is spent where it buys the most; novelty is reserved for adaptive music and configurations.
Collapsed subgraph node with clean exposed ports
FIG. 04 Subgraphs: complexity folds into a node with a clean port contract. Same canvas one level deeper, so the concept count stays low and teams share proven behaviors.
Patch configuration switching parameter profiles without duplicating the graph
FIG. 05 Configurations: one graph, many parameter profiles. Kills the copy-the-patch anti-pattern that silently forks logic.
Arrangement view with instrument tracks and segments
FIG. 06 The arrangement: deliberately DAW-shaped for composers. Variation, motifs and stingers are authored objects, not code behind the music.
Profiler window showing runtime voices and performance next to authoring
FIG. 07 The profiler: runtime made observable next to authoring. We positioned ELIAS against black-box middleware, so this claim had to be rendered as UI.
04 · Outcomes & takeaways

What the design work unlocked.

Funding

The design work was a core asset in the successful funding round.

Team scale

Design grew from one designer to a dedicated 5-person team building inside the system I established.

Foundations

IA, core flows, the node editor interaction model, and the design system shipped as the durable skeleton of the product.

Open debt

Graph search and a minimap were still open design debt when I left; at hundreds of nodes, find and jump matter as much as editing.

Outcomes as credited during my tenure; funding specifics are the company's confidential information.

What I'd reuse anywhere. Pick one loop and collapse it. Spend familiarity to buy novelty. Make the data model visible, so users debug themselves. And give pros escape hatches: extensibility isn't a power feature, it's a trust feature. ELIAS 4 is challenger UX in one line: don't out-feature the incumbents, out-flow them.

Screens © Elias Software / House of Elias.

Next case study

Aligni: right-sizing PLM and MRP

Read next All work