Clara Smith is a Technical Audio Designer who lives in Unreal Engine’s native audio tools: MetaSounds, Blueprints, the works. She recently wrapped a run at Epic Games on Fortnite live content, and before that she worked on a fast, anime-styled first-person shooter built entirely on Unreal’s native audio.
She didn’t bring slides. She brought a demo project, posted it to GitHub ahead of time, and spent the hour inside it with a room of us following along.
Native Audio Is the Engine
The line Clara most wanted people to leave with: native audio in Unreal is the engine. Not only MetaSound graphs. It’s MetaSounds plus Blueprint plus visual effects and animation, which means the technical audio designer ends up in everyone’s graphs, and every Slack channel, because the music needs everyone’s data.
MetaSounds themselves are DSP (digital signal processing) graphs. If modular synths mean nothing to you, think of Unreal’s material graphs instead. A Blueprint runs left to right, one command after another. A MetaSound graph doesn’t. The whole thing evaluates at once, every audio buffer.
That one fact explains a lot of beginner frustration. Why won’t Unreal let you wire a wave player’s On Finished straight back into its Play? Because you’re asking it to stop and start on the same sample, and a MetaSound can’t be in two places at once.
Native audio in Unreal is the engine.
— Clara Smith
Build for Two Potatoes Stuck Together
Fortnite has to run on phones and the Nintendo Switch. Clara’s shorthand was blunter: make sure the potatoes can run it, even if it’s two potatoes stuck together.
So performance becomes a design input rather than a cleanup pass. Mega graphs (one MetaSound doing everything) are the first warning sign. Wave players are the biggest memory cost on the graph, and her rule of thumb for multi-platform work is about six per graph, tops. Her demo ran at five: one for stingers, plus two “decks” so one phase can fade out while the next fades in, the way a DJ crossfades.
Small things add up as well. Every extra variable instance costs something, while reroute nodes cost nothing. Nested graphs are tidy and modular, and each extra level of nesting loads another graph and its nodes, so she keeps it to one level where she can.
Limit yourself to about six wave players, tops.
— Clara Smith
The sneaky one is volume. Turn sounds down at the sound class and the engine’s voice management sees it, then virtualizes or culls anything too quiet to hear. Turn them down on a submix and the engine still thinks every sound is playing at full volume. Free performance, if you know where to put the knob.
Let the Music Listen to the Game
Her demo had enemies spawning into an arena and an energy meter in the corner that climbed as they arrived. More energy, more percussion. Music in games, as Clara put it, gives emotional context, spatial awareness and gameplay cues, and combat intensity is the easy place to start.
The simple version was two layers playing together, with the energy layer’s volume rising as the fight grew. One catch: enemies arrive whole, so the value jumps in steps. She smoothed each jump over a quarter note at the track’s tempo, which kills the clicky transitions.
The advanced version used phases, the method she used on Fortnite. Phase zero is idle, phase one is first contact, and by phase three the choir comes in. Transitions wait for the next beat so the groove survives. Percussion lives on its own stems, separate from everything else, because kicks and snares fight weapon fire for the same frequencies. Whenever a weapon fires, it ducks the percussion (audio people call it sidechaining).
Somewhere in the middle of all that she said the thing every overthinking developer needs to hear. There’s no right way. The only way that is correct is shipped.
Don't overthink it, there's no right way. The only way that is correct is shipped.
— Clara Smith
Things to Keep in Mind
- Start with footsteps. Clara calls them the hello world of tech audio.
- Write an audio design doc before you build anything, covering your north stars, your scope, and the systems you are deliberately not building.
- Export music and effects as 48 kHz, 16-bit WAV files, Unreal’s default, so the engine has nothing to convert.
- Put the BPM in every music file name. Your metronome needs it, and you won’t have to keep asking the composer.
- Use Spawn Sound 2D instead of Play Sound whenever you need to control the music later. Only the spawn version hands you an audio component.
- Keep your music manager somewhere that survives a level change, like the game instance or a game instance subsystem. A world subsystem loses your music when the level unloads.
- Move game logic out of MetaSounds. Phase tracking belongs in Blueprint or C++, and the graph should just take the result.
- Turn volume down at the sound class, not the submix, so voice management can do its job.
Get Audio to the Table Early
Questions brought out the best story of the night. For Halloween 2025, the Fortnite team had four weeks to rework an existing game mode into Demon Rush, the mode built for the KPop Demon Hunters crossover. Clara did all the audio programming for it, and got to build something that normally never happens: music that carried on from the battle bus all the way to the ground, where players usually drop in silence.
She also had a warning for small studios. Audio is always last in the pipeline, and audio finds your bugs. Half the time, she said, she was the one spotting gameplay bugs, because a sound played when it shouldn’t have.
How did she learn all this? YouTube tutorials, game jams that left her no choice, and pulling apart Epic’s Lyra sample project node by node with breakpoints.
If you missed it, the recording is coming soon to the OGDA member portal for Associate members. Clara also said she’d happily come back for a hands-on workshop, laptops open, getting footsteps in. If you’d come to that, tell us in Discord.
If you can get your audio team at the table early, you will save yourself headache in the long run, because audio finds your bugs.
— Clara Smith
October 1st: James Smith
The talk after Clara’s swapped audio for operations. James Smith, CEO of BlackCap Creative and one of OGDA‘s founders, presented The Studio Survival Paradox: Why Operations Make Better Games on October 1st.
His subject was the next stage. Once developers have pitched, networked and landed funding, what comes after? James calls that moment the success trap: buy-in arrives, momentum spikes, and the trip you planned to walk suddenly has to be driven. His fixes started small. If a new feature means rebuilding your framework, it waits for the next game. And spread operational know-how across the team, so nobody sits idle until one person gets back.