The Video Game Composer's Problem

Video game music had a particular argument for AI-resistance that seemed persuasive through about 2023. The argument ran like this: game audio is interactive. A game score must respond to player actions, game state, and dynamic conditions in real time — shifting between combat and exploration, transitioning smoothly between areas, adapting to procedurally generated events. This interactivity requires not just composition but technical implementation: understanding middleware like Wwise and FMOD, building music systems that respond to game parameters, working closely with audio engineers and game designers. AI couldn’t do all that.

The argument was half right. AI cannot fully replace the music systems implementation work. What it could do — once it could do it at all — was generate the raw audio content that these systems play. And that, it turned out, was enough.

The Indie Game Transformation

The disruption hit indie game development first and most severely, which should not be surprising. Indie developers work with constrained budgets where every line item is under pressure. A full original score for a mid-sized indie game cost $15,000-$40,000 in composer fees in 2022. An equivalent amount of AI-generated game music that could be licensed for a flat fee of $200-$500 from emerging game audio AI services was not compositionally identical — but it was sufficient for the game’s commercial purposes.

By 2025, several of the most successful indie games on Steam used entirely or primarily AI-generated music. Depths of Aether, an exploration game that sold over 400,000 copies, used a Stable Audio-derived system to generate adaptive exploration music dynamically during gameplay — a procedural approach where the music was literally generated in real time based on game parameters. There was no score in the traditional sense. There was no composer. The developer, a team of three, had a soundtrack that adapted to the player’s position and emotional context more responsively than any pre-composed score could have.

The response from the game composer community was, understandably, unhappy. The Game Audio Network Guild issued statements. Several composers publicly declined to work on projects that had also used AI audio tools. The ethical arguments were genuinely made and genuinely inadequate as economic solutions: the developers who used AI weren’t making a philosophical statement, they were solving a budget problem.

The AAA Situation Is Different

At the large budget end of game development — the AAA games with budgets of $50-200 million — the situation is more complex and the disruption has been slower and more partial.

Several major studios have announced or confirmed the use of AI audio tools for specific applications: procedural ambient soundscapes, NPC dialogue variation, dynamic adaptive systems that complement a human-composed score. None of the major AAA releases of 2025-2026 has publicly claimed to use AI for their primary musical scores, and the composers of those scores — the Austin Wintory and Jesper Kyd tier — remain human.

What has changed at the AAA level is not the primary composition but the composition support work: the placeholder music used during development before the final score is recorded, the variations and stems generated for adaptive music systems, the incidental music for minor areas that might previously have warranted a brief composer engagement. This work, which was previously contracted to mid-level composers and studios, is now largely AI-generated in-house.

The mid-level game composer — competent, reliable, not a brand name, sustainable as a business — has lost significant work. Not the senior composers who bring identity to major titles, and not the entry-level composers whose cost was already minimal. The middle.

The Adaptive Music Problem No One Solved

There is a genuine technical frontier that AI has not yet closed: truly adaptive music that responds to emotional game states in ways that are musically sophisticated.

The best examples of adaptive game music — the Wwise-implemented systems in games like Control or Disco Elysium or Red Dead Redemption 2 — work by transitioning between pre-composed musical stems according to a sophisticated state machine. A composer designs multiple versions of each section (combat, exploration, tense, resolved) and the system selects and transitions between them based on game state. The quality of the result depends entirely on the quality of the compositional decisions: where to cut, how to transition, what emotional quality to use for each state.

AI systems that generate music procedurally are not good at this yet. They can generate consistent ambient textures. They struggle with the kind of dramatic arc that makes game music feel responsive — the moment when the musical tension releases in synchrony with a game event, the swelling theme that coincides with a narrative revelation. These moments require a composer who has played the game and understands its emotional structure, not just its audio parameters.

This gap is narrowing. Audio AI research is specifically targeting adaptive music generation, and the timeline until AI can generate musically coherent adaptive scores is shorter than most composers want to believe. But it is not zero yet, and the window matters.

What Survives

The most economically durable position for a game composer in 2027 is as an integrated creative partner rather than a pure music deliverable provider.

Composers who are embedded in game development teams — who influence the emotional arc of the game’s design, who consult on how music systems should be built, who bring a recognizable aesthetic identity to a title — are in a different position than composers who are hired after the game is largely finished to write music to fit existing scenes. The former is harder to replace because it involves creative partnership with the development team; the latter is exactly the task that AI is well-positioned for.

The composers who have retained or grown their practices in the AI era are mostly those who have built reputational identities strong enough that hiring them is itself a value proposition — audiences for games by composers with strong identities will seek out those games partly for the music. This is a small cohort. It does not provide an answer for the larger population of working game composers.

The video game composer’s problem is a version of the general creative industries problem: AI disruption is not uniform across prestige tiers, and the middle is where it lands hardest. The artistic quality argument is true but insufficient. A mid-level composer who makes music that is better than AI-generated music is still in trouble if the client cannot charge enough for their game to justify the difference in cost. That is the real problem, and it is not solved by making better music.