It was the obvious first question to ask. Was the answer going to be to create highly realistic audio, or should we mimic sounds and use substitution for using real samples? Perhaps sound design can't even contribute to create a fantasy mood and instead music holds the answer.

Thinking about such a broad question made obvious there was not going to be a straightforward answer so instead of aiming for a high concept solution and perhaps getting lost in the weeds, the attention was directed at pragmatic problems such as restrictions imposed by platform, budget, milestones, game-play, audience, technology, art direction. Then look towards similar games for inspiration. Breaking down the unknowns into knows, so the problem becomes less daunting.

Understanding a projects imposed restrictions will help when defining scope and setting boundaries for making creative decisions quickly and in a focused way. That might seem like confining creative freedom but it is in fact a liberating step. Knowing the goal helps focus decisions in the right direction and maximizes usable creativity. In my experience this is one of the more difficult learnings from (game)-development and many of the creative people I admire most seem to find this effortless. I'd be lying if I said I didn't still struggle with it at times.

My fascination for sonic journeys (ambient, film music and Lifeforms by The Future Sound of London specifically) played a role in setting a creative direction. By that I mean, to deploy all the tools in our sonic arsenal (music, sound design, voice) to tell stories; Find the detail in the world and the moods portrayed within it, then put a lens on those with sound design. With technical, budget and time limitations understood, I eventually settled on dense streaming audio soundscapes, augmented by limited diegetic spot sound and a (short) music score that would highlight achievements rather than serve as continual aural wallpaper (or plaster even).

The act of walking through the environment in The Chronicles of Spellborn would be akin got creating an ever changing, impressionistic audio environment. It influenced the design of audio, including voices and music and every technical audio decision can be traced back to this origin. It became the first pillar for audio design: describe the world not just as it appears, but also the place it takes in the lore, stories and quests. From this first pillar a second became apparent; To describe the world, the moment to moment details of traversing needs to be expressed in an always changing soundscape; how else to convey movement if the sound remains more or less static?

With a creative direction in place and the project's scope charted, I had a grounding focal point to work from.

A concept drawing of a game world (shard) taken from the vision document.

While researching games released around the same time (in 2005) I noticed that many had used a singular ambience covering relatively large environments. While sparse spot effects were used to represent specific occurrences (a stream of water or a campfire) what I found lacking was granular transitions to really convey the idea of moving from one area to the next.
Almost all games I listened to relied on music (often looping) to set the mood and I reasoned it to be an understandably effective and resource friendly approach to paint audio onto a large scale open-world setting.

Developing games is hard enough as it is, so following tried and tested method is a sensible strategy. It was however at odds with the creative audio pillars. How could one ambience or piece of music effectively communicate a sense of progression or transition between similar yet different acoustic environments? A forest might be a forest, but it will sound different on an open dirt road or when under a canopy of leaves. One generic ambience could not take the many possible detailed transitions into account.

Village, Fields, Forest and Demon Wall. Music can be used to tell the story of the overall area while specific environmental ambiences transtion between all these sub-area.

A lesson learned from being a level designer on Alpha Black Zero and Eurocops is that the environment can be an important narrative tool in a first and third person game. For instance, showing a goal at the start of a level provides narrative focus ( go there ) and level designers still use visual cues acting as a reminder for experience, progression and grounding direction.
Think about the collapsed high rises at the start of The Last of Us or the giant Erdtree in Eldenring. In the later game, each area has the equivalent of a larger structure towering above the area and giving players a direction where to go. Though where it goes wrong for some players is the third campfire pointing towards Stormveil Castle when most players cannot proceed there without becoming stronger necessitating traveling South away from the castle, instead of North. Some level designers are masochists indeed.
So mostly it works well because in an interactive medium the actions of the protagonist are controlled instead of observed and progression through an environment is a story told at a players own pace. In an interactive medium we don't just identify with a character, we control their very actions. Sound can become a narrative tool by associating it to smaller parts of a larger environment in such a way that transitions occur when moving from one area to the next.

In games, everything is an approximation. In the real world, sound, reflections and propagation are continuously changing even when we do as little as slightly move our heads. How detailed a game can get is determined by two factors: The amount of time available to create the audio and the technical processing & memory available to render it in real-time. The smaller a described area, the closer we can get to a "real" experience. From initial research at the start of the project, I knew that our tech, time and budget restrictions would not make it possible to create unique sounds for every possible transition.

I did wonder if this same conclusion had stumped other audio departments while reaching this point in production? Was their pragmatic solution to rely on atmosphere and music instead of trying to paint more intricate transitions? In other words, did they choose to implement audio based on the progression of (linear) time out of practicality, rather than sound ignoring to acknowledge a players choice to move in any given direction?

The MMORPG's of the time that I tested would use three layers of audio for their environments:

• A limited amount of spot-effects for specific objects in the environment (expensive on real-time render and memory resources)
• An ambience to give life to a larger landmass (cheap to store and play back and it can be hugely detailed)
• Music to set the mood and convey emotion

There's a practical back-to-front approach there: Music conveying emotion. Ambiences convey the general mood, time-of-day or environmental sounds and spot-sound-effects supporting in world objects that require it. Technically those are different systems with their own requirements for processing time and memory: Spot-effects have a location in the world and become louder when moving closer to them. Ambience and music can be streaming loops that are triggered to play or stop by game events. One resides in world space and memory and related directly to nearness. Streaming loops are restricted to the time-domain and can be influenced by game data though will in essence always 'play' forwards. The advantages are obvious, streaming required bandwidth and system memory, while location based sound require processing and need to reside in memory. This often reduces their quality and length as often competes for memory with other departments such as art and animation. Using streaming loops works well for MMORPG's precisely because of the much smaller footprint while allowing much more detailed audio and its relative abstraction from the actual game environment ensures that there this is always something to be heard. It however cannot express the movement and progression of a player as closely as I would wanted to achieve.

Several experiments, prototypes, meetings and tests later we settled on a 'best of both worlds' approach: Use streaming loops (can be long, convey loads of detail and are high quality) while using in world makers (with falloff) to set the volume informed by an in-world location. A bit like shuffling a playlist of music, but instead of it happening randomly at the end of each song, we could do it based on location. We already had point-source sounds with a radius and an in world position (comes with Unreal 2.5) so coder Brian Wren re-purposed it to trigger audio loops and set their volume accordingly. The streaming ambiences allow for detail and depth as their length is only restricted by storage space not memory. Brian's to be commended for his contributions as it didn't require radical new tech, instead his solutions expanded upon tried and tested methods.

Top: a city environment as it appears in game. Below: displaying the 4000+ "sound-sources" (as green wireframe spheres) that define the audio for all the seperate areas.

If movement through a large world requires many different sounds to represent that world as it moves past, limited memory and processing power put a cap on the maximum amount of sounds to be individually placed in that world to achieve this. Our solution as discussed was to package a majority of detail into streams. A crude analogy for this is that you can transport more marbles with one hand when you put them into a bag first and then transport the bag instead of each individual marble.

Johan Cruyff (famous Dutch footballer and Coach) once said 'Every advantage comes with a disadvantage'. The main drawback to streamed media (in the way expressed here, this is less an issue with modern faster storage) is that having only a few of them cannot be made sufficiently interactive to player movement. The individual sounds rendered into a stream can't be re-arranged or repositioned based on the movement of a player through the world.

If I wanted for sound to follow transitions between smaller areas of an environment (small enough to be noticed anyway) we would need a way to inform audio where the player is going in the world and define those areas to link them to a corresponding ambiences. Spot sounds already do this, but they are expensive when deployed in great numbers. Audio coder Brian Wrenn used that functionality to inform nearness to a locations manipulating the volume of an ambience. I would have to carefully limit the amount of overlapping ambiences to prevent streams taxing the limited hardware.

Our solution was to create virtual speakers and link those to virtual audio players. The speakers are "virtual" because the sound isn't actually coming from them, they do however inform nearness and function to specify locations in the world. When a player 'walks' inside a virtual speaker, the linked virtual audio player begins streaming audio and plays it back at a volume proportionally to the distance from the center of the in-world sphere. The player is 'virtual' because it doesn't do so for that location, it simply plays it back like any music player software on the computer would.

Virtual speakers (circular wireframes) are linked to audio players (icons at the bottom). The sound from the audio players is heard when within the sphere, loudest at the center and attenuated towards the edge.

When sphere's with different streaming audio overlap, moving between them gives the impression of the sound changing gradually according to that location thus unlinking it from passing of time. Think of it as if having two speakers in a room playing different content; The closer to either speaker the louder the audio will sound, or said differently, as you travel through the room, the sound changes.

The virtual audio-players could be configured to play-back any kind of sound (directional effects or stereo streams) and Brian added extra functionality to modulate volume, pitch and interval over time. These days that functionality is exposed through game audio middleware such as WWISE of Fmod but we had design and build it into Unreal 2.5 ourselves. Brian and me had a blast coming up with more features as he made the streaming system more and more efficient.

The approach was usable because a single virtual audio player could be hooked up to thousands of virtual speakers without ever stressing the system. It wasn't actually playing thousands of instances, merely using nearness as data to drive the loudness of the stream. Modern middleware such as wwise refers to this as a multi-point implementation (multiple emitters, one decoded stream). A player could never occupy all virtual speakers at the same time (unless they are all in the same spot) so moving between was little more than starting/stopping a stream and raising or lowering the volume. This was not new tech, but we added a layer of innovation to make it reactive and scale well for our worlds. This also wasn't entirely new, as trigger based starting and stopping was a function of Unreal 2.5. We innovated on how this trigger worked to achieve a denser, richer audio experience.

You may have already guessed that special care had to be taken when placing spheres as to prevent excessive overlapping. The system did allow us to place a magical "sparkling" sound for hundreds of lamp-posts, add the sound of leaves rustling to all the trees in a map or individual insect sounds near specific plants. Within days an entire area could be populated with sounds as it was quick to play back and intuitive to place. When moving past objects, the sound changed accordingly. It was quick and relatively cheap to produce, but immersive and interactive to the experience. More important, it was not possible with a one to one (traditional) way of implementing and added depth to the audio experience of Spellborn as remarked in many reviews.

For an average sized area:

• Roughly 150 virtual audio players for playing back:
  • streaming audio (cheap/detailed/compressed audio)
  • directional sound effects (expensive/mono/uncompressed)
  • ambiences fill in the soundscape of the world while directional sound effects detail it
• An average map contains some 2000 virtual speakers per area
• Almost all streams needed a day/night variation to accommodate for the day/night cycle, doubling the amount of streams needed during transitions between day and night

Brians' optimisation kept audio within an average 1ms per rendered frame, very little if when considering a frame can take up to 33ms to render a 30fps game. If audio is 30% of the experience (as quoted by Steven Spielberg) we sure weren't achieving it at that cost to processing, though did shift some of the cost onto storage.

Using streams, the audio in The Chronicles of Spellborn matched our audio pillars and to some extend exceed expectations from initially ambitious goals: It was indeed possible to recognize an area and even sub-areas from the sounds heard within it and convey the sensation of movement, even with eyes closed.

A heavily modified version of the Unreal 2 shows 135 Virtual Audio Player objects floating at the edge of the Hawksmouth map. A selection of the spheres is highlighted. Note also the 'elongated' shapes that Brian added to compliment spheres. He also created a ring emitter (donut shaped). Hawskmouth contains 1311 virtual audio speakers. We didn't introduce square shaped emitters for their impact on performance. Showing a single virtual audio player connected to multiple spheres to describe those areas. The Red sphere is a "hotspot" in which the volume doesn't attenuated.

Having solved some of the technical hurdles, the solution of using streams was a compromise, as is often the case in game-development. Going back to the earlier example of the marbles in a container, that method of transportation locks them into pre-set configurations. The red marbles can't move towards the yellow ones, without opening that container. Equally, in a stereo stream a sound that is rendered to the left will always be heard as coming from the left, regardless if someone in the game world is running around in circles. This could invalidate the illusion of depth and movement that was so carefully woven into the creation of the linear ambience streams.

After months of testing and fine-tuning I came to a remarkably simple approach: Much like the brain can be tricked into perceiving a rapid series of "frames" as a fluid motion our ears can be similarly fooled into perceiving depth.

Adding directional spot-sounds in the world to complement the ambience tricked ears into believing everything heard had direction. The trick was to make sure the sound in the spot overlapped enough with sounds in the ambience. The effect turned out to be so strong that one or two carefully chosen sounds would be enough to achieve an overall depth of perception. As directional effects are much more expensive needing only a few was a workable compromise.

The effect was convincing enough that we never once received a beta-player or co-worker comment which had noticed almost the entirety of sound in The Chronicles of Spellborn did not move to the left or right as a player turns round. To the contrary, we even received comments that the world sounded real; '... like playing with the windows open ... '. That was the best compliment imaginable. I had experimented with more sophisticated solutions like rotating part of the stereo stream or creating quad ambiences but results did not justify the higher cost in processing and memory usage. Sometimes,realism can be achieved by not directly mimicking reality.

Instead of the impossible route of using (too) many directional spot-sounds to narrate transitions or designing complicated technical approach for interactivity in audio streams; We achieved our goals through iteration on existing systems. The end result was equal or sometimes even exceeded comparable games from the time, while it didn't force our players to invest in new sound hardware (which as anyone working in audio knows, doesn't exist to begin with).

An infinite amount of sounds, effects and music can be packed into a compressed, streaming ambience that is 'placed' at a specific location in the world.

With a technical solution, tools and direction sorted, it was time to think of creating the ambiences themselves. This was informed by the second audio pillar in that audio should put an emphasis on change. Meaning that even linear streams could communicate a sense of progression. As established; Movement through the world translated to crossfading between different ambiences.

To take this a step further, the streaming ambiences also required a progression for when people were not moving between different ambiences. They would otherwise hear the same sound of wind or birds over and over again. In a real forest birds move around all the time and the wind is usually not constant either.

To achieve progression in ambiences I applied principles learned in 15 years of creating ambient environment soundscapes. What makes (good) ambient music unique is that it strives to entertain without asking for attention. It can slip into the background while still being interesting enough when listened to in a focused way. When applied to sound design the ambient approach helps to mask repetition and prevents audio from becoming predictable or worse; boring!

The ambiences for The Chronicles of Spellborn are all structured like an evolving piece of ambient music to tell stories so that the environment appears to be alive. Here's a short representation of an ambience for the starting area. Layers weave in and out, often panned around the center to create a sense of movement and wind going through the world. The birds, crickets and tonal elements are reactive to these layers and fade around them, further conveying the sense of wind actually influencing the fauna.

An infinite amount of sounds, effects and music can be packed into a compressed, streaming ambience that is 'placed' at a specific location in the world.

Notes from the audio design document: Sounds of the wind fade in and out. Use filtering, effects and panning to have the appear as to come from far away and move towards the listener. As the wind move through trees, rustle leaves in the distance before the wind engulfs the player in a swift gust. Theis creates swells that ebb and flow around the listener, constantly changing the sound of the ambience. Quiet the layers of insect and bird song to create a "chilled" atmosphere. A bird flies up in distress. As the wind calms, chimes sound somewhere in the distance. Bring back the warmth using a minimal synthesizer/tonal ambience. Once the mood feels warmer again, fade in the layers of cricket sounds and birds.
For storytelling, an infinite amount of sounds, effects and music can be packed into a compressed, streaming ambience that is 'placed' at a specific location in the world. Which would be hard to replicate using diegetic (in-world) spot sounds.

The following audio/video example demonstrates this progression:

From field to docks - using tonal and composed layers to create a sense of progression in ambiences, even when standing still. Audio is deliberately hard-cut between locations (as it played there) to highlight how it is different both in time passed and location. The field ambience follows the principles of ambient music, interesting enough for someone that is actively listening yet it doesn't impose itself when playing the game.

From experimenting two minutes is the sweet spot between storage used and repetition. Shorter and the loop became obvious and much longer only increased the overall file size of the game. In placing ambiences I always took into account that players would need less than a minute to move from at least one ambience (area) to another.

On average every ambience was built from a collection of 15 to 150 individual sounds. Outdoor areas required a day and night version to be created and follow that cycle in the game. Ambience's would cross-fade between both versions at dawn and dusk over a short period of time to minimise stream doubling.

Considering the transitions between day and night and the possibility of music playing, there would always be a minimum of three streams active in any single (outdoor) location. To make the game as accessible as possible, I restricted the amount of overlapping sphere's containing streaming audio to a maximum of three. Although Unreal's optimized decoder and Brians' code could easily handle the decompression of fifteen simultaneous streams before drop outs, to remain on the safe side, having three spheres restricts the maximum simultaneous streams to seven. Not problem in confined linear areas, but in open areas it was often a puzzle when implementing audio and trying to make it all fit.

To prevent repetition fatigue I approach ambiences in frequently visited areas different from those in more unique locations. A frequent location would have a transparent ambience whereas unique locations could contain more outspoken audio, such as whispering, musical elements or sharper contrasts.

On average an area contained eight different ambiences though more than twenty was not an exception. Over 250 ambiences where created which equals a total play time of over 15 hours.

Dutch national radio 4 recognized the audio achievement and it broad-casted a one hour special using ambiences, with voice recordings of Anna Drijver and musical compositions by composer Jesper Kyd. It is linked at the top of this page.

Creating sound-effects can be a straight forward process: Wind can be recorded outside or taken from a collection of ready-made libraries, even synthesized using plugins and hardware.

But what does a Howler sound like? Such sounds can't be recorded outside and are created by layering existing sounds and guided by the imagination.

Streaming ambiences would be created to cover audio environments, but as mentioned before, sound effects would be required for adding depth-perception to those ambiences. As a sound designer one rule I will always follow is that when something can be seen moving, it should probably have a sound to re-enforce that.

Process: Audio is created to a linear video capture. Created audio is exported into sub-components (footsteps, Foley, voice etc.) and extra variations are made where necessary. The groups of sounds are then linked to specific frames in the animation. The sound effects for creatures where grouped into sub-topics like breath, fur, footstep, idle noise, scream etc. On specific frames in a creature's animation those groups where linked for playback. Various frames in an animation could trigger a random sound with random pitch and volume applied, creating varied, layered but still recognizable feedback for movement, attack and defense. Animation Tool: Triggers for sounds are added to frames in a loaded animation. Each trigger links to a sound and a defines distance range. This allowed important sounds to be heard far away, while details remained for close-ups. This made creatures sound organic as the combination of overlapping sound-effect and volume, + pitch variations created an ever changing soundscape and loads of detailing and depth. It worked especially well when creatures where in large groups. I carefully used the ranges to allow an area to be populated by sounds of the creatures inhabiting it. Being able to spot them well beyond their aggressive ranges for some of the larger sounds, while being able to add loads of detail for close-ups. This created a blance between storytelling and keeping creature audio performant. Creature Effects Reel - Starts playing the Howler Flagelant from the examples above. More recognizable creatures such as boars, bears and rabbits also made an appearance in The Chronicles of Spellborn. We divided them into two groups. For predators we used sampled recordings and for prey and critters we used members of the staff to voice them. Pitched up and processed many devs on the team became creatures in the game.

The predator AI would actively hunt down non-aggressive creatures and people alike so this devision made both more distinct and recognisable. It also made the critters sound incredibly cute, which was exactly what we needed to make the predators appear more beastly by contrast.

The spell-room that contains visual (particle) effects for skills, combat and spells grouped into schools of magic and combat. In Spellborn combat revolves around buffing and de-buffing over time. Using a unique combat wheel mechanic that required players to sort skills into rows of sequential combo's it was essential to time 'firing' of skills and being close enough to hit. That's unusual for an MMORPG even in 2025 and combat was uniquely active and played more like a FPS game: real-time and required manual targeting and dodging.

As movement and timing required more concentration, the combat sound-effects would have to provide feedback about the type of magic being used amd what buffs where being applied.

As fighters concentrated on aiming and dodging attacks they had little time to read combat-logs or status updates. Spellborn shipped with close to 200 skills, grouped into school of magic and combat, each group received a distinct acoustic theme. We prioritised readability over form and indeed a group of players could unleash many different spells simultaneously yet it was still possible to pick out what schools of magic or combat where being used.

Spell Effects Reel - Starts playing various spellcasting effects.

A unique selling point of The Chronicles of Spellborn was freedom of choice. This was especially apparent in being able to create completely unique characters and skill-decks. We also decoupled gear from stats. I applied a level of choice to dialogue also. in a few new directions with audio also. As players were allowed to dress any way they wanted (instead of being restricted to class or type of character) the choice of voice was also up to the player. We created eight different archetypes such as hesitant, aggressive, wise, leader and recorded the same expressions with different actors in that archtype exposing the choice of voice (and associated 'type' during character creation.

I wrote the initial lines and -- let's be kind and say -- they sounded very much written by a non-native speaker. Professional writer Helen Brady wrote a second pass and put her background in theater and novels to great use. She rewrote most of the text adding flair and a bit of that typical British dry wit. Our players loved this personal approach and some of them would meet up to exchange greetings and emotes just to hear all the unique lines.

The actors where cast and recorded under the supervision of casting director Darren Scott who's experience in writing and directing stage plays quickly managed performances out of the actors. On a relative small recording budget we were able to create distinct and interesting dialogue.

Each part had an average of 300 sentences, sounds and grunts. We had to avoid expressions such as "good day" as we didn't record with an awareness for night or day time (or any other eventually as such). Of the 48 actors, 16 where selected for users to pick during character creation. The rest were used exclusively for Non Player Characters. The npc's were also recorded using character traits so that game designers could assign specific voices and tones to suit the stories they had written. One exception was made for the voices for bosses who received the same voices as players; this made them appear more dangerous.

Top: Libby McKay, huffing, puffing and screaming at the top of her lungs. Middle: Anna Drijver, voice of narrator and protagonist in the lore. Bottom: Voice Director Darren Scott and engineer and studio owner John Rivers at Woodbine Street Studios [UK] We settled on using stage actors because of their ability to infuse personality into parts. Working with actors who have played contemporary and classic stage plays brought a level of authenticity. achieve. Darren Scott suggested that action parts were recorded last. Ten minutes of screaming and grunting would wear voices out.

The art of The Chronicles of Spellborn mimics a hand drawn style inspired by that of quaint Dutch artist Anton Pieck. A fullr traditional orchestral score might have be too rich for the style of the visuals.

After months of listening to composers and creating a short list, it was almost by accident that we found the perfect match in Jesper Kyd. The music in Io Interactive's Freedom Fighters caught our attention. Freedom Fighters presented a blend of electronics, inventiveness and acoustic sounds and it was obvious that Jesper had an incredibly talented and gifted musical voice.

Before choosing him, we selected two composers and paid them for a demo. Eventually, the talent and experience of Jesper Kyd won out and his demo presented and already fully fledged Spellborn theme. On his website Jesper (at the time) he wrote that the music was '...his most affecting and diverse score to date...'.

Taking our time to find the right composer paid off as Jesper Kyd's style of music was a natural fit to The Chronicles of Spellborn. We spend hours discussing the setting of the world, stories and environments and when he delivered his music, it was always spot on.

Jesper Kyd understands the sensibilities of games and music and how they affect each other. I initially opted to reserve his music for precious moments in the game only. A decision I had to reverse thanks to the many beta testers who requested that his music be heard more often. What greater compliment could he have received?

Jesper Kyd was the right choice for The Chronicles of Spellborn and I cannot imagine any other music to have fit the world so well.

Complete Spellborn Soundtrack - All the music composed by Jesper Kyd plus my own tonal ambiences. These are harvested from the game release so lack the in game volume adjustments.