The boss dies, the music cuts, and your heal fires into a scene that suddenly has nothing else in it. The cue you approved under a wall of drums now sounds thin. There is a click at the head you never noticed, and the reverb keeps ringing after the health bar is full.
Nothing about the cue changed. The context did. Most heal sounds are approved in combat and then spend their life in the quiet after it.
Designing for the quiet is a separate pass, not a volume trim. It means re-checking level, tail, noise floor and duration against the sounds that remain.
The Mix You Approved Is Not the Mix That Plays
In combat, a heal cue is balanced against a busy bed. Music might sit near -14 LUFS, impacts peak around -6 dBFS, and the heal peaks near -12 dBFS. It reads because the ear has a reference.
Remove the music and the reference disappears. A -12 dBFS cue is now the loudest thing in a scene that may sit at -55 dBFS. Nothing was boosted, but the perceived change can run 6 to 12 dB, past the 3 dB where listeners start calling something louder.
The fader is often the wrong tool. Turn the heal down until it sits under a boss roar and it vanishes in the quiet, which is where it spends most of its runtime.
Music and impacts are combat-only. Footsteps, interface sounds, ambience and the heal are always present. Balance the heal against that second group and let the combat bus duck it when needed: a sidechain on the music bus, an AudioMixer snapshot in Unity, or an RTPC-driven gain in Wwise. The mechanism matters less than the reference. A practical test is to mute the music bus for one lap and note which sounds you can still hear.
The Tail Now Has a Room to Fill
Reverb is a masking problem during a fight and an audible object after it. An 800 ms tail at -20 dB is buried under gunfire and unmistakable in the silence that follows.
A long hall decay reads as continuation, which suits a channelled heal that takes 1.5 seconds to land. It fights an instant heal that resolves in 120 ms, because the ear hears the cue finish and then hears it keep going. The mismatch usually surfaces in playtests as a vague sense that healing is unreliable.
Match the tail to the space, then decide whether it should imply more healing. Non-diegetic confirmation sounds often work better dryer than the room suggests. How the tail shapes recovery goes further into that trade-off.
Two checks catch most problems. Shorten the tail until the cue ends before the next likely event, then high-pass the body at 300 Hz and listen to the tail alone. A resonant ring will be obvious in the quiet and probably masked in combat, which is why it survived review. Solo the tail 6 dB below normal to hear the ring without the body masking it.
Room Tone Is an Exposer
Combat hides artefacts. A 2 ms click at the head, a DC offset that thumps on trigger, a hard cut at the end of a reverb: all of it disappears under impacts and sounds unfinished alone.
Export format matters. Keep sources at 48 kHz, 24-bit WAV and do level work in floating point. If the engine compresses to Vorbis or Opus, check the shortest transient, because heavy compression can soften a 5 ms attack, and the attack is most of what makes a heal read. Render a 16-bit reference copy too, because some builds still ship that.
Then check the noise floor. An ambient loop at -50 dBFS is inaudible under combat and clearly present in the quiet. If the heal was rendered at a different floor, a hiss that stops when the cue ends points straight at the seam.
Raise the monitor level 12 dB above normal and listen twice. Once for artefacts, once for anything that reads as a second event: a secondary resonance, a delay tap, a flutter in the tail. Fix what you hear, then confirm the fix did not soften the cue.
In a No-Music Game, the Heal and the Footsteps Are the Soundtrack
A game without a score leans on repetition, and its two most frequent sounds set the rhythm of the experience. Footsteps might land every 0.4 to 0.5 seconds. A heal might fire every 20 to 40 seconds.
If the heal shares a band with the footsteps, the player hears a slightly louder footstep and no confirmation. Separate them by spectrum. Footsteps hold weight between 100 and 800 Hz and scuff between 2 and 6 kHz, so put the heal identity near 500 to 800 Hz with harmonics falling away above 4 kHz.
Level is the second lever. A workable hierarchy is damage above heal, heal above footsteps, footsteps above ambience: a hit peaking at -6 dBFS, a heal at -12 to -15, footsteps at -18 to -22. These are starting points. A game with no music often tolerates a louder heal, because nothing else competes for that role.
The opposite failure is filling the silence too aggressively. Filling silence without overdoing it covers the wider balance. The heal-specific version is to accept that a sparse scene makes every cue louder than its fader suggests.
Short and Dry, or Long and Wet
Follow the mechanic, not the genre. An instant heal and a two-second channel need different envelopes, and the cue is often the only timing information the player gets.
Short and dry
Instant heals, pickup-style heals and regeneration ticks usually want 80 to 180 ms of body, an attack between 5 and 15 ms, and a release closing inside 250 ms, with little or no reverb. That keeps the cue out of the way of repeated triggers and stops a 40-second cadence from building into a wash. A dry cue also stacks better: two heals 300 ms apart read as two events, while two wet cues read as one smear.
Long and wet
Channelled heals, safe-room restores and narrative recovery moments can carry a 1.5 to 2.5 second tail and a body that rises instead of landing. The envelope can hold a sustain for the channel and release when the heal completes, which ties the sound to the mechanic. Keep the wet signal below the dry so the cue still has an edge.
A middle option suits many games: a dry confirmation followed by a short room tail. The dry part answers the question, and the tail matches the space without implying the heal is still in progress.
The Check That Matters Is Against Footsteps, Not the Boss
Most heal cues are auditioned against the loudest thing in the game. That is the least useful comparison available. The cue will live next to footsteps, doors and ambience.
Build a test scene with no music, a walking loop at the real cadence, and the heal firing at the real interval. Listen once at normal level, once at half volume on a phone speaker, and once with the heal pitched down a semitone. If the cue disappears, the problem is usually spectral overlap rather than absolute level.
Timing is the other half. A heal triggered on the frame the health value changes is usually sample-accurate through the audio engine, but a cue started from a script that also spawns a particle effect can drift by a frame or more. At 60 fps that is 16.7 ms; a 100 ms trigger delay is not tolerable for an instant heal.
The last check is the pair test. Play the damage cue and the heal cue back to back and ask whether they are clearly opposite. If both rise, or both share the same noise transient, the quiet mix exposes the similarity combat was hiding. Why heal and damage cues must be opposite covers the pairing logic; health pickup cue design has the shorter cue-level constraints.
None of this requires a new sound. It requires a second mix pass in the context where the cue actually plays, with the music muted and the footsteps running.