Roblox Audio Graph Editor
Dynamic audio tooling to elevate sound in gaming experiences.
Role
Product Design Intern
Timeline
May - Aug 2025
Team
1 Designer
1 Engineer
Tool Kit
Figma, Cursor,
Claude Design
01 SUMMARY
Improving the engineer experience with analytics for frictionless operations
As the sole designer on the Audio Graph Editor, I built a no-code tool helping technical artists among Roblox's 1.3 million creators build game audio, which drives the experience, accessibility, and story cues. Creators often left Studio to build audio in other engines, and those who used our Audio API relied on engineers to code it by hand. While my work is under NDA, my main contributions included:
As the sole designer on the Audio Graph Editor, I built a no-code tool helping technical artists among Roblox's 1.3 million creators build game audio, which drives the experience, accessibility, and story cues. Creators often left Studio to build audio in other engines, and those who used our Audio API relied on engineers to code it by hand. While my work is under NDA, my main contributions included:
Designed and shipped the Audio Graph Editor, making basic audio setup 3x faster (3 minutes to 1) with 62% fewer clicks by turning a code-heavy API into a visual, no-code node graph
Designed and shipped the Audio Graph Editor, making basic audio setup 3x faster (3 minutes to 1) with 62% fewer clicks by turning a code-heavy API into a visual, no-code node graph
Cut node validation time ~60% across a 22-node system by rapid-prototyping interaction variants with AI in Cursor, running 5 iteration cycles after just 2 weeks of research with zero audio background
Cut node validation time ~60% across a 22-node system by rapid-prototyping interaction variants with AI in Cursor, running 5 iteration cycles after just 2 weeks of research with zero audio background
Expanded Roblox's Foundation UI design system from outside the team, creating 6 new Studio-specific components that led Foundation UI to build native graph support for other teams
Expanded Roblox's Foundation UI design system from outside the team, creating 6 new Studio-specific components that led Foundation UI to build native graph support for other teams
03 TAKEAWAYS
Design for technical users' mental models
Technical users like sound designers and audio engineers think in systems, not screens. They want to see how everything connects, then experiment freely within it. Matching that mental model, instead of splitting work into separate views, reduced context switching and cognitive load, and gave them room to explore.
Technical users like sound designers and audio engineers think in systems, not screens. They want to see how everything connects, then experiment freely within it. Matching that mental model, instead of splitting work into separate views, reduced context switching and cognitive load, and gave them room to explore.
Treat constraints as deliberate tradeoffs
A technical constraint meant breaking a convention users knew from related tools, which risked consistency and standards. Rather than force the pattern, I found a lightweight way to preserve the information architecture users relied on. Being transparent about what we gave up, and why, mattered as much as the solution.
A technical constraint meant breaking a convention users knew from related tools, which risked consistency and standards. Rather than force the pattern, I found a lightweight way to preserve the information architecture users relied on. Being transparent about what we gave up, and why, mattered as much as the solution.
Iterate fast, and document every decision
As the sole designer, I learned to trust my work and pivot quickly. AI-assisted prototyping let me bring several iterations to each crit, each with its tradeoffs and design rationale written down, including why I rejected the alternatives. When my design validator left the company, I found another, and my decision log meant I could answer "why not this?" with evidence instead of starting over.
As the sole designer, I learned to trust my work and pivot quickly. AI-assisted prototyping let me bring several iterations to each crit, each with its tradeoffs and design rationale written down, including why I rejected the alternatives. When my design validator left the company, I found another, and my decision log meant I could answer "why not this?" with evidence instead of starting over.
Any questions? Contact me at dkherrer@uci.edu to see the full case study!
Any questions? Contact me at dkherrer@uci.edu to see the full case study!
Any questions? Contact me at dkherrer@uci.edu to see the full case study!
