Guest Contribution by Michael Theiler of Kpow Audio
We always knew The Banner Saga was going to be something special. First contact with the guys behind it was so positive and just easy. Our meetings and emails hit all the rights notes creatively so much so we just knew it was going to be an exhilarating ride. We weren’t wrong. Working on a project with such rich creativity and depth behind it, paired with the encouragement and trust the team showed, buoyed us, and heightened our commitment to the project.
There were a few key areas we concentrated on when it came to the sound design of The Banner Saga. We wanted to pay particular attention to the ambiences of the locations in the game – to make them real and evocative and never distract the player from the game. We wanted to make sure the banner, which is always with you but changes in length depending on how the player is progressing, had gravitas and importance without overwhelming the player. Likewise for the scenes where travelling by cart is depicted, the cart sounds were incredibly important as they were the sounds of the population traversing vast distances, fleeing from a dark force. We needed the fighting sounds to be gritty and real, and for their special abilities to have a different sound to them. We wanted the fighters to feel like they were pulling their strength and concentration for their actions from an internal well of ancient power. Finally we wanted the UI to not draw too much attention to itself, to feel solid and real but distinct from the other in-game sounds.
Ambient Audio
I have written previously about the creation of the ambient audio for The Banner Saga: Factions, which is the free multi-player game created before the single player version that I am writing about here. The ambient audio in the single player version were similarly crafted, and used many of the same sounds, but they were much larger in scope, and more varied. We used a fairly common system now, whereby you make hundreds of small sounds, and get your audio middleware (in this case FMOD) to play them back randomly, with playback rates and times meticulously tuned so that in concert all these little sounds begin to feel like a real ambient space.
To achieve this we started off by deciding what the main beds were going to be in the various locations. Usually they were different qualities of wind events sitting on an ‘air’ event. The air events were the glue that held all the ambient audio together, the base element. The wind events were then added on top of these. We had seven different wind events that could be mixed and matched for different locations. The main wind events were made up of between 10 and 17 samples. These samples were created with fade in and out times between two and seven seconds long, so that they blended seamlessly with the ambient air base events. They were triggered to play a sample every 0.7 seconds and 3.7 seconds, and could have up to three samples per event playing at once. This meant that usually at least two samples of wind were playing per wind event, which created our gently moving sound bed.
Each wind sample was designed to have subtle panning, by manually panning elements across the soundfield, and by flipping the stereo on around half the samples so that their stereo elements would feel evenly balanced when played back together. The combination of the implementation and the design of the individual samples meant we had a rich, deep, always moving and never the same ambient bed. Combined with the air events, we then had a great base for all the other ambient elements.
To add highlights and interest to the ambience, we had a number of audio events that were used during the travel sequences and in the various locations the player visits. These were all setup in a similar way to wind, but were triggered far less frequently. They consisted of sounds such as water lapping on shore, ice cracking, distant waves, seagulls, forest leaves, forest shrubs, forest birds, many walla events, many distant foley events for cloth, foots, metal banging, flag-poles etc, interior sounds such as mugs on tables, chair scrapes, wood creaks, and many many more.
Setting up ambient audio in this way, while it can be difficult on an organizational level, allows for the ambience to feel very organic once balanced correctly for all locations. This was an important element to the game, as to a large degree, the environment is also a character. We needed for the world to feel like it had depth – it was important for the experience. Even though the ambient audio is often played fairly low in the background, there are still benefits in having such variety and lack of repetition.
The Banner
To ensure there was also a lack of repetition in the banner sounds, we set up the banner event in a similar way. Each banner event was comprised of multiple layers, which could be faded in and on based on wind strength and camera position, creating a flapping, moving, always different sounding large banner.
To get these sounds we had a recording session in a newly built studio. Having the luxury of a recording session for a relatively small title was great, but due to the sound we were trying to capture and the newness of the studio, we did run into some issues. The sound of cloth flapping and waving in the air is basically broadband noise, so any peculiarities in the room become exacerbated and apparent in the recording. I had brought a lot of sound dampening panels to the studio to help alleviate any unwanted reverberance in the space, but not enough, and we were hearing the reflections of the room in some of the takes. We adjusted the panels and mic positions and got useable material that was made to sound ok after processing later on, but if I were doing it again, for large cloth flaps and fabric movement, I would try to get into a much larger space, with better treatment.
The good thing about getting out and recording for this vital element of the game was that we now had a heap of useable and varied flapping banner and flag sounds. Once edited and added to the banner event we had 24 individual whips for the banner whips layer. This layer was for the short sharp whips you get on strong fast gusts of wind. We had 14 samples for the banner natural layer, which were longer movement sounds that played quite often and had a rather large pitch randomization on them. Finally we had the banner slow layer, made up of 23 samples and contributed to the sound of longer, more languid snake movements of the banner. Combined into the one event, these were then made to sound different by effects which were controlled by parameters that were altered based on distance of banner to player, and by wind speed. These parameters altered the layer’s volume, spawn intensity, and pitch, all differently for each layer.
We put a lot of energy into the banner as it was such an important part of the story the character wove as they progressed through the world. Its size represents the size of your caravan. As you are immersed in this world, the significance of the banner grows. Its an organic seeping – there are only a handful of story points that fill in its significance. But the movement of it, and its alterations in size as you progress, these all lend it a sense of importance. It too is a character, and one we needed to get right.
Travel Sounds
Travel in The Banner Saga is made up of a long shot of a caravan, composed of various amounts of peasants, fighters and Varl, all marching through various glorious vistas. We created two events for the caravan – one for close and one for distant. The sounds of the caravan were made up of many quite small sounds playing back randomly, much like our other generative events, but on a more granular level. They had a variance parameter which gave subtle pitch variation quite slowly over time, as well as reducing the level of the event slowly over time so it never gets too annoying. The other parameter is for the speed of the cart, and alters the volume level of the layers in the cart event as the cart speeds up or slows down. Having these two separate parameters both affect the volume level and pitch of the event means the cart always sounds different. It pitches up when it first starts, which indicates the strain of acceleration. It then levels out and the volume start to imperceptibly drop so the ambience can come through and music isn’t obscured. The pitch variation on the variance parameter ensures the pitch movement of the cart varies every time, so you don’t notice a particular pattern to its pitch changes.
The sounds for the base of the cart, and for wood rattle of the cart are setup on two separate layers. In addition, the cart event has the sound of the ox which is triggered only very occasionally, and the sound of marching feet, which is fairly subtle, but ensures the caravan feels populated.
Together with the banner sounds, these events make up much of the sounds associated with the characters you are controlling. As you travel throughout the unbelievably beautiful scenery, these are the sounds that accompany your caravan, and therefor represent another important sound to get right, while not drawing too much attention to them.
Fighting Sounds
Much of the fighting in The Banner Saga is with traditional weapons. Axes, swords, maces etc. We split the impact sounds for weapons into two sdi’s, which would give more variety in weapon impacts, and allow for mixing and matching of samples. We had the body of the hit in one sdi, and the slice or stab on another. We could then vary the intensity of the body sound depending on the character being hit. We could also share the body sounds across many different weapons, and just change whether the other sound was a slice for sword, or a crunch of armor etc for mace. Getting a system like this setup correctly for all the different armor types requires careful organization and good communication, but ultimately gives more variety in weapon sounds.
Many characters also have an active special ability – one that can be activated by the player during a battle. These are actions such as a roundabout swirl of a sword, or a bow and arrow shot that goes through multiple targets, or flurry of axe hits impacting randomly strength or armor. For these special abilities, we wanted the sound to be the power of the special, and come from a deep, ancient place. We wanted it to be the focus of the character, channeling this old power from within themselves. To achieve this we tried to get a similar growing distant texture happening in the background of all the special moves. They are all slightly different, but all have a spacey, swirling quality that grows into the action of the ability. Where all other battle sounds are mono center, we gave these sounds a lot of width, and used subtle delays and spacialization techniques to ebb and flow these sounds out of the background and into the ability. The sounds are not too over the top, and are not your usual fantasy magic type sounds. We hope they retain some animalism and primal qualities, which we felt was important to the setting.
User Interface
The ambient audio fills out the game world, with the one-shot ambient sounds making it feel more alive, lively and inhabited. The banner, along with the caravan sounds makes the party you are leading across the snow more tangible and real, while the battle sounds are gritty and visceral, lending more immediacy and drama to the battles. The User Interface had to sit on top of this, and communicate to the player what they were doing, when an action is taken, and they had to do it without drawing too much attention to the fact that it is an artifice.
In order to place the UI sounds in their own layer of audio in the world, they were given a different specialisation treatment. All sounds were eq’d too be sure they didn’t muddy up the lower end of the eq spectrum, but this was carefully done so the sounds still had weight to them, as we wanted them to have an old, slightly chunky, slightly magical quality. Like gold metal and crystal sounds rather than fantasy magic sounds. We used a plate reverb for all the UI sounds, and a short stereo delay with slight variation in slap timings between left and right. Using a plate delay for UI can be a bit heavy handed and naff, but we tried to keep it fairly subtle so it gives a different sense of space to all the other sounds, but still gives a sense of place.
The sound we were going for was imagining the UI all coming from an old wood room filled with books and parchment and implements, while being a little bit pagan magical. Its a fine line. The source material for the UI sounds were varied, but all a bit similar. Metal chain chinks, clasps of wooden chests, tibetan singing bowls, crystal glass ringing and chinking, and bone and feather type sounds were the main contenders, and created the design for which we were aiming.
The overall design for the audio of The Banner Saga came together as well as we could have hoped. We comprehensively put into practice many generative methods for creating sound design in game that we have been developing, while keeping it all controlled and specific, servicing its particular area. Compartmentalising all the sound design elements so that they suit their role and maintain their own pocket or space in the overall sound of the game is something that we hold to be incredibly important. Of course Austin Wintory’s score elevated the whole experience to another level, and working with him was incredible. The entire team were amazing, and we couldn’t have hoped for a better result.
Kpow Audio – currently seeking projects
Michael Theiler – LinkedIn profile
Kpow Sounds – we have some sound libraries available from our webstore
Richard Gould says
Lovely breakdown, I especially enjoyed your approach to distinguishing UI sounds. Looks like you made powerful yet sensitive use of the tools, it’s given me food for thought for a project which I’m working on at the moment. Many thanks and congratulations on the IGF Honorary Mention in the audio category, well deserved indeed.
Michael Theiler says
Thanks Richard, the nomination meant a lot, though Austin Wintory’s score probably had a lot to do with it!
Dave Chow says
Thank you so much for revealing your designs behind what I found to be a gorgeous and engaging sound experience.
Michael Theiler says
Thanks Dave, I’m glad you enjoyed it so much!
rene coronado says
I did a little youtube search for some gameplay, and there’s a great little bit of it here:
https://www.youtube.com/watch?v=fq790mfIjto
Jump to 5:10 seconds in to hear the excellent ambiences referenced up top.
UI sounds are there from the beginning. Killer work all around.
Michael Theiler says
Thanks so much Rene, and thanks for tracking down a youtube video without a voice over!
João Costa says
Great read! I´m new to FMOD so could you please explain what do you mean by SDI?
Michael Theiler says
Hi João,
SDI is the shortened term for Sound Definition, a container for audio files which can have their parameters tuned (replays, time between replays, pitch randomization etc).
João Costa says
Hey Michael,
Thanks for the reply. Cheers