Video summary
Next Level Creature Sound Design
Main summary
Key takeaways
Summary: “Next Level Creature Sound Design” (GDC 2017)
David Philipp (Noise Works) presents a practical framework for designing believable creature sound effects for games, focusing on how to create lifelike “one creature” vocal behavior without real-world references.
Speaker background / credibility
- Co-founder of Noise Works Limited (audio outsourcing), London
- Boom Library recorder and stem designer
- Creature-sound experience in titles like:
- Elder Scrolls Online
- Blade & Soul
- Arcania
- Total War: Warhammer (led creature SFX production)
Core concept: “Believability” through character, variation, and perspective
Philipp frames great creature sound design as three “ingredients”:
-
Character
- How the creature acts/emotes: aggression, intelligence, stupidity, emotion
- Vocal basis: attack, death, pain, idle behaviors
-
Variation
- Variation in articulations (attack/death/pain/idle)
- Variation in sound sources and recording techniques
-
Perspective
- Spatial placement: indoor/outdoor, cave/close-up/far away
- Drives how the sound “lives” in the game world
Goal: produce something that isn’t real, but is believable to the player—supporting immersion.
Pre-production: planning before recording
Philipp emphasizes pre-production as often neglected but crucial.
Key steps include:
- Learn the creature’s origin, special skills, and characteristics
- e.g., dragon fire breath, tail swing, “mega scream” buffs allies
- Determine how the creature produces sound
- whether it has a voice box / vocal mechanism / non-human anatomy
- this affects vocal approach and what articulations are feasible
- Confirm scale/size relationships
- difficult to fix after recording full sets (small vs massive creatures)
- Plan a sonic “map” (mind-map style) and timeline
- example: planning a random dragon to a total of 42 sound effects
- Decide on potential sources from:
- humans
- animals
- non-literal recordings (objects/props)
Recording process: microphones and source strategies
Philipp argues that different microphones capture useful layers for pitching and blending.
Microphones highlighted
-
Sanken C100K
- Captures up to ~100 kHz (useful for pitching without losing detail)
- Pitch-down trials: record raw then pitch down multiple octaves while retaining crispness
- Notes:
- omnidirectional → harder outdoors
- stops working below ~10°C
- workaround: wait for warmer weather / use pocket warmers + tape/rigging
-
Sennheiser MKH series
- Field-friendly; commonly used
- ~half frequency response vs C100K (~50 kHz), still usable for pitching
- Available in cardioid / supercardioid (e.g., MKH 8014), often used as MS setup
-
Contact mics (Barcus Berry)
- Captures dry, isolated signals (throat, objects, room elements)
- Expensive; higher-end models provide higher fidelity vs cheaper options
- Used as alternate “layers” that later become part of creature texture
Human recordings
-
Pros
- Directional and can perform “on command”
- Cheaper than animals (especially without actors)
- Easier to record in controlled environments
- Can request “weird” sounds from coworkers via email (example workflow)
-
Cons
- After pitching down (2–3 octaves), humans can sound similar and messy
- Sore throat/hoarseness limits how far performers can go
-
Best practice
- Maintain variation through different recording techniques, not just pitch/speed changes.
Animal recordings
-
Pros
- Some sounds are hard/impossible to reproduce convincingly with humans
- Natural character (horses, bears, seals, vultures, etc.)
-
Cons
- Animals often won’t vocalize on cue; field time is unpredictable
- Uncontrolled environments create heavy cleaning/editing work
- Costs can be high if agencies train animals for specific performances
- Some setups can be dangerous (mating season, wildcats, stags, etc.)
-
Field tips
- Choose quiet locations (example: UK near flight paths/motorways)
- Record surrounding ambience first (build an archive)
- Talk to animal handlers/trainers for “tricks” (e.g., vocalize when hungry)
- Plan insurance / safety; boom setups accordingly
Non-literal recordings (props/objects)
Philipp calls this the most interesting and creatively challenging technique.
-
Examples
- tire squeals, balloon scrapes, door creaks, metal scrapes, dryers on metal
-
Advice
- “look for sounds, not props”
- carry a recorder / notepad and capture unexpected everyday noises
-
Demonstration idea
- Using balloon scrape layers plus a human scream can create a creature-like composite with new character.
Design phase: layering and the “homogenous creature” requirement
Creature sound design is fundamentally layer-based, but the craft is making layers behave like one creature.
Key design principles
-
Internal tonal relationship matters more than plugins
- Pitch matching between layers is critical
- EQ/volume offsets help balance harmonic/body resonance
-
Pitch matching example
- Adding a new layer can disrupt tonal harmony → requires re-pitching other layers
-
Analogy
- Like arranging music: a guitar riff and bass must be in tune.
Tooling notes: plugins used (processing support)
Examples mentioned for processing include:
- iZotope RX / RX features
- denoise, declip, cleaning
- Synaptic Morph 2
- morph spectral characteristics of one sound into another
- Elastic Pitch
- automating pitch over time (e.g., pitch shifts when a creature dies/falls)
- Waves / (Magique Q mention)
- uses EQ knobs to restore brilliance/highs after heavy pitching loss
- Wave/other “bigness” shaping plugin
- Dehumanizer 2
- large parameter set, modular, automatable; integrated into Pro Tools via plugin format
Plugins support the work, but believability depends primarily on layer/source relationships.
Implementation: game integration is “50%” of success
Philipp emphasizes that even perfect audio can fail if integration is wrong.
Game audio system tips
-
Hierarchy/setup matters
- Example: Total War: Warhammer used creature sizes: small/medium/large/extra large/huge
- ensures big creatures stay big in-game mixing
-
Modular vs one-shot articulation
- One-shot: attach a trigger to an animation event
- Modular: split sounds (e.g., scream into start/middle/stop) for tighter animation synchronization
-
Perspective and distance workflow
- Start with close-up sound layers
- Also create distant versions (often derived from close layers by removing layers to simulate distance)
- Blend by in-engine parameters
-
Cone attenuation / directional EQ
- adjust lowpass/highpass based on creature facing angle and camera direction
-
Gain staging
- large creatures should be louder than small ones
- recommend setting initial gain offsets to make mixing easier
-
Sweetener/proximity layers
- add extra stereo/quad “proximity” textures when creatures are very close
Takeaways / checklist style guidance
- Pre-production is key
- understand creature mechanics, skills, scale, and sound production method
- Record as much source material as possible
- build a library for variation and layering options
- Design with the foundational trio
- pitch matching, EQ, volume
- Don’t rely on “the plugins”
- they support the work, but source and layer relationships determine believability
- Get involved in implementation
- or coordinate tightly with implementation/engine/audio teams
- Final goal
- character + believability for immersion
Main speaker / sources
- David Philipp — co-founder, Noise Works Limited; boom library recorder/stem designer; GDC 2017 presenter (primary source of the talk)