Chapter 30 — Key Takeaways

1. Run the mute test on your own game. Monthly. Play with sound off and see what is comprehensible. Then play with sound on and see what audio is doing. The gap between the two is the value your audio is providing — or, if there is no gap, the work your audio is failing to do. This is the most useful audio diagnostic you have, and it is free. Most designers never run it on their own work. Be the exception. The first time you do, you will discover that audio is doing dramatically more work than you thought, and you will find specific sounds that need fixing. The tenth time, you will hear games at a level your peers cannot.

2. There are four audio layers. Treat them as first-class design concerns. SFX (instantaneous events), Music (emotional spine, minute-scale), Voice (line-scale dialogue and barks), Ambient (area-scale world beds). Each has its own bus, its own volume slider, its own design concerns. A designer who writes "audio: slashing sound" in their spec is shipping a game with audio problems already baked in. Spec audio with the same care you spec visuals — what is the source, what is the layering, what is the variance, what is the spatialization, how does it interact with the mix.

3. SFX are the audio layer of game feel. Layer them, vary them, time them right. A great hit sound is not one sample — it is an attack-impact-body-release envelope built from layered samples (metallic + meaty + crack), with pitch variance to defeat repetition, fired on the anticipation frame so it lands with the visual contact. Footsteps need ±10–20 percent pitch variance and a pool of 3–5 alternates. Any sound played more than four times in a row needs randomization. Treat each SFX as an engineered artifact, not a sourced file.

4. Music does at least four jobs. Composers who hit all four are the ones to hire. Music sets mood, paces energy, signals transitions, and builds identity. A composer who only thinks about mood (film-scoring habit) misses pacing and transition work. A composer who only thinks about themes (theme-writer habit) writes inflexible music. The best game composers — Wintory, Larkin, Raine, Gordon — hit all four because they understand the medium's specific demands.

5. Pick the simplest dynamic music technique that meets your need. Usually a state machine. Vertical layering (stems) is great if you have a composer who can write parallel-track music and an engineer who can fade them in real time. Horizontal re-sequencing (iMUSE-style) is great if your composer can write segments that connect at musical boundaries. State machines (cross-fade between named modes — explore, combat, boss) are simpler, do most of the perceptible work of the others, and are what you should build first. Implementation in Godot: three AudioStreamPlayers started simultaneously and tweened on volume.

6. Voice is the most expensive layer. Be honest about what you can afford. Full VO is AAA. Major-character VO is mid-budget. Bark-only VO is achievable for many indies. No VO is a legitimate artistic choice (Undertale, Hollow Knight) provided you design the text-delivery system with care. Decide your VO tier early; designing as if you will have full VO and then cutting back is more expensive than starting with bark-only and adding later. When you do hire actors, hire for interpretation and direct seriously — fifty great-acted lines beat five hundred workmanlike reads.

7. Ambient makes worlds feel real. Layer it, never single-loop it. A single forest loop sounds great for 30 seconds and increasingly artificial after that. The right pattern is layered ambient: a continuous base bed, a secondary bed of randomized one-shots (birds, drips), a tertiary bed of distant elements (wind, distant water), and occasional one-shot interjections at random intervals. The brain does not pattern-match a layered bed because there is no single pattern. Alien: Isolation does this masterfully and uses the ambient as gameplay (steam vents mask movement; quiet rooms reveal it).

8. Diegetic vs. non-diegetic is a real design choice. Be deliberate about it. Diegetic audio exists in the world (a radio, a band, a phone). Non-diegetic exists for the player (the score). GTA's radio stations, BioShock's vinyl recordings, and Cadence of Hyrule's beat-locked mechanics all break the convention deliberately. Most games use a stable mix (non-diegetic score + diegetic environmental and dialogue). When you cross the line, do it for a reason. When you stay on one side, do that for a reason too.

9. Mix is the designer's responsibility. Do not delegate it away. Music ducking under voice, dynamic-range compression for varied playback environments, frequency separation between layers, stereo positioning of spatial sources — these are all decisions about what the player hears moment-to-moment, which is to say they are design decisions. Mix on multiple playback systems (studio headphones, laptop speakers, a phone). Target -18 to -24 LUFS integrated loudness. Implement automatic ducking from day one, not as a final-week scramble. A great game with a bad mix ships with a footnote in every review.

10. Accessibility is not optional. Five settings buy you the entire conversation. Separate sliders for Master / Music / SFX / Voice / Ambient. A Subtitles toggle that captions both dialogue and gameplay-critical SFX. A Mono Audio toggle that sums to mono for single-ear players. Vibration as an audio proxy for deaf players. None of these is hard to implement; all of them are mandatory in 2026. Audio-driven games (horror, stealth, atmospheric experience) need more accessibility work, not less, because their gameplay depends on the channel that is impaired for some players.

11. Budget audio is a learnable craft. Hire composers, DIY SFX. Music composition is hard to fake; a passable original score beats a polished library track, and serious composers will work on indie projects for fixed fees or revenue share. SFX work is engineering as much as art — Freesound plus Reaper plus a phone-recorded foley pass produces a respectable indie SFX palette. Original-recorded SFX in your closet are often better than library samples because they are yours. Spend a week learning a DAW; it pays back forever.

12. Build the audio layer alongside everything else, not after. The temptation in production is to ship gameplay-complete and add audio "in polish." This is a disaster. Audio bugs found at the end are baked into the systems they touch. Implement AudioManager.gd early. Wire it into every system as you build that system. Run the mute test monthly. By the time you are in polish, your audio should already be 80 percent there. Polish is for tuning, not for first-pass.

These twelve principles will not produce a great audio layer on their own. They will, however, prevent the most common audio failures and give you a framework for thinking about audio as design. Apply them in order. Run the mute test before you run the next one. Your project, and the next ten years of your career, will be better for it.