Many games strive for realism. They aim to reproduce the sound of the world as accurately as possible, carefully modelling spaces, materials, physics, and behaviours so that players feel immersed in a believable environment. Don’t Starve takes a rather different approach. Its world is filled with living scarecrows, walking trees, giant spiders, impossible creatures, and surreal landscapes that seem to have escaped from the pages of a dark storybook. Very little about it appears realistic in any conventional sense. Yet despite this, the game feels remarkably alive.
Matthew Marteinsson’s guest lecture explored how that happened. Although the talk covered specific technical systems, recording techniques, production challenges, and implementation details, a broader idea repeatedly emerged beneath them. The sounds of Don’t Starve do not feel convincing because they imitate reality. They feel convincing because they remain connected to physical performance, playful experimentation, and a constant willingness to explore unexpected possibilities.
During the early development of Don’t Starve, Marteinsson was effectively the sole audio designer working alongside two composers, with no dedicated audio programmer and no substantial audio department behind him. Development moved rapidly, content changed constantly, and there was little opportunity for elaborate production pipelines. Rather than treating these limitations as obstacles, the team repeatedly used them as opportunities to find simpler and more creative solutions. Constraints were not merely something to overcome. They actively shaped the character of the game’s sound world.
The game’s character voices provide an excellent example. Traditional voice acting would have required large quantities of dialogue recording, scheduling actors, and continuously updating content as the game evolved. Such an approach was difficult to reconcile with the speed at which the project was being developed. Yet the characters still needed personality, emotional expression, and identities that players could immediately distinguish. Instead of using spoken language, Marteinsson turned to musical instruments. Inspired partly by the adults in Peanuts cartoons and partly by Peter and the Wolf, where different instruments represent different characters, each character in Don’t Starvereceived its own instrumental voice. Wilson’s distinctive muted trumpet became the starting point, with subsequent characters developing from their own carefully chosen instrumental identities. What began as a practical solution ultimately became one of the most recognisable features of the game.
Human vocal performance appeared repeatedly as a creative tool throughout the lecture. Many memorable sounds originated not from extensive libraries or complex synthesis chains but from experimentation with the voice itself. The spiders, for example, were largely built from Marteinsson’s own vocal performances combined with processing. The Gobbler, one of the game’s most beloved creatures, began with attempts to gather suitable turkey recordings. After examining the animation, however, he found himself instinctively making a strange vocal sound that immediately felt more appropriate than any authentic turkey call. The library recordings were discarded and the vocal performance became the creature. As he noted, the deliberately exaggerated human performance communicated personality far more effectively than realism alone could have achieved.
Realism and believability emerged as distinct ideas within Marteinsson’s approach to sound design. A perfectly accurate turkey recording might have sounded more realistic, though it may not have felt more alive. The Gobbler succeeds precisely because it occupies an unusual space between animal, caricature, and performance. Players are not simply hearing a creature. They are hearing a performance of a creature. The sound communicates character as much as biology.
Personality often seemed more important than realism throughout the lecture. Many of the creatures in Don’t Starveexist within a visual world that is intentionally exaggerated, stylised, and slightly absurd. Conventional fantasy sound design might have felt strangely out of place. Marteinsson instead described grounding many creatures in a “weird reality”, where recognisable physical behaviours remain present but become filtered through performance, humour, and experimentation. Human vocalisations proved especially valuable in this regard. Audiences are extraordinarily sensitive to nuances in human expression. Even heavily processed vocal sounds can communicate intention, emotion, vulnerability, aggression, or curiosity in ways that are difficult to achieve through purely synthetic or animal-based recordings.
Environmental audio presented an equally interesting challenge. Procedurally generated worlds create difficulties that traditional environmental sound design rarely encounters. Designers cannot assume where players will travel or which environments they will encounter. Marteinsson described a system that continuously examines the terrain surrounding the player, identifies the dominant biome types within the immediate area, and dynamically blends the corresponding ambiences. Grasslands, forests, marshes, and other environments continuously mix together according to what the player is actually seeing at that moment. Rather than creating a fixed soundtrack for a predetermined world, the system responds to the world being generated in real time.
What makes this system particularly interesting is that the underlying idea remains remarkably simple. Players should hear the world they are looking at. Technical sophistication only becomes valuable when it strengthens the player’s experience. Systems matter not because they are complex but because they help players understand the world around them. Throughout the lecture, Marteinsson repeatedly demonstrated a preference for elegant solutions that serve a clear experiential purpose.
The broader design philosophy became especially clear during the discussion following the lecture. Marteinsson argued that game audio should generally perform one of two functions: it should either build the world or inform the player. If a sound accomplishes neither, its value becomes questionable. Such a statement sounds straightforward, though it carries considerable implications for design practice. Many games accumulate audio over time, layering additional sounds onto already crowded mixes. The result can be confusion rather than clarity. Marteinsson instead advocates careful consideration of why a sound exists and what purpose it serves. Sound is not decoration. It is communication.
Small details often became surprisingly important within this design philosophy. During the lecture, he discussed how player feedback during early access revealed complaints about a particular pickup sound. Some players even requested a dedicated option to disable it. Rather than immediately changing the sound itself, Marteinsson investigated further and discovered that the underlying issue was simply that the sound was mixed too loudly. Once its level was adjusted, the complaints disappeared. The lesson was not that players were wrong. Rather, it highlighted the importance of identifying the underlying problem rather than accepting proposed solutions at face value. Players are often very effective at identifying areas where something feels wrong. Determining why it feels wrong remains part of the designer’s responsibility.
Recording sessions often sounded closer to scientific experiments than conventional sound production. Music boxes, improvised instruments, jelly, pudding, toys, unusual household objects, mines, and novelty items discovered in shops all found their way into Marteinsson’s recording collection. A music box originally intended for composing melodies eventually became the basis for the unsettling sounds associated with the Shadow Hand. A visit to a local mining museum produced unique underground ambience recordings for the game’s cave systems. Strange objects were collected not because a specific project required them, but because they might become useful in the future.
Playfulness often appeared not as a break from the work but as part of the work itself. Making strange noises while watching an animation, experimenting with unusual objects, collecting sounds without a specific purpose in mind, or exploring unexpected combinations of recordings all reflect a willingness to follow curiosity wherever it leads. Listening to these stories, it became increasingly clear that creativity often depends upon creating opportunities for surprise. The value of an unusual object or recording does not necessarily become apparent immediately. A sound designer may encounter something intriguing, record it, store it away, and only discover its purpose years later.
Technical decisions rarely appeared separate from creative ones during the lecture. Recording techniques, implementation systems, middleware, debugging tools, and production constraints were all discussed in detail. Yet none of these elements were treated as separate from creativity itself. Debug tools existed to facilitate experimentation. Procedural systems existed to strengthen immersion. Recording techniques existed to discover new forms of expression. Technology remained important throughout the talk, though it rarely appeared as the primary source of innovation.
Reflections on game development brought many of the lecture’s themes together. Marteinsson acknowledged the challenges facing the industry, including long hours, instability, and periods of significant uncertainty. Yet his reflections consistently returned to enthusiasm, curiosity, and the joy of creating experiences that players genuinely care about. That optimism felt closely connected to the ideas that had surfaced throughout the lecture. The sounds of Don’t Starve emerged not from a search for perfection but from a willingness to experiment, adapt, collaborate, and occasionally embrace absurd ideas simply to see where they might lead.
Perhaps that helps explain why the world of Don’t Starve feels so distinctive. Its sounds rarely seem trapped by expectations about what things ought to sound like. A spider may begin as a human vocal performance. A terrifying shadow creature may emerge from a modified music box. An iconic turkey may owe more to an impulsive noise made while watching an animation than to any field recording. Throughout the lecture, Marteinsson repeatedly demonstrated that memorable sound design often emerges when curiosity is allowed to guide the process.
Rather than attempting to recreate reality exactly, Don’t Starve constructs a world that feels alive through performance, experimentation, and play. Many of the sounds discussed during the lecture began as accidents, improvisations, constraints, or strange ideas that simply seemed worth exploring. What emerged from that process was not merely a collection of sound effects but a coherent sonic world. Listening to the lecture, it became difficult to separate the sound of Don’t Starve from the spirit in which it was created. Both are defined by curiosity.
In doing so, Marteinsson offered a useful reminder that some of the most memorable sounds are not discovered by following established rules. They emerge when designers remain willing to ask a simple question: what happens if we try this?