Score & Sound Design: Scoring Emotion
Score & Sound
Music licensing has kept the film industry and the music industry in productive conflict since synchronisation rights were formalised. AI music generation breaks the model at both ends: it eliminates the need to license existing music, and it produces original-sounding compositions that may or may not be copyrightable. The lawsuits are in progress.
Universal Music Group Sues Suno and Udio for Copyright Infringement — Claims Training Data Included Full Recordings
Universal Music Group, Sony Music, and Warner Music jointly filed suit against AI music platforms Suno and Udio, alleging that their models were trained on copyrighted recordings without licence and that the outputs reproduce protected elements. The suits seek damages of $150,000 per work infringed — potentially the largest copyright case in music history by value.
"The number is theoretical. The principle is not."
Hans Zimmer Argues AI Cannot Write "The Silence Between Notes" — Releases Composition Tool Built Collaboratively With AI
"Music is time," Zimmer said at Venice. "AI generates notes. It does not yet understand what it means to decide to wait." Two months later, he released a composer tool built with Sony AI that assists with orchestration and harmonic suggestion while leaving tempo, silence, and emotional arc to human composers.
"He is right about the limitation. He is also building the tool that might solve it."
Spotify AI DJ Feature Displaces Human Playlist Curation — 2.4 Million Playlists Deleted as Traffic Falls 61%
Spotify data showed that the launch of its AI DJ feature correlated with a 61% drop in user-generated playlist creation and millions of deletions of existing playlists. Independent playlist curators who had built audience followings through Spotify's editorial programme saw follower counts collapse as algorithmic curation replaced human-authored playlists in user behaviour.
Film Composers Guild: 34% of Members Report AI-Generated Score Drafts Used as Starting Points — Without Credit
A survey of 1,100 professional film composers found that 34% had been presented with AI-generated music in a "starting point" or "reference track" capacity by production companies, often in productions where the AI contribution was not disclosed in the contract. In 19% of cases, the final score bore structural similarities to the AI draft.
If an AI music model trained on 10 million copyrighted songs produces a composition that sounds stylistically similar to but note-for-note different from any of them, has any infringement occurred?
Music And Emotion
Music shapes emotion through established patterns your brain recognizes. When a major chord rises, your nervous system interprets it as uplift or optimism. When a minor chord descends with sparse instrumentation, it reads as introspection or loss. This isn't cultural magic—it's pattern recognition. Your auditory cortex processes frequency relationships (the math of intervals), your emotional processing centre links those patterns to learned associations (minor = sadness because you've heard it paired with sadness thousands of times), and your body responds with physical tension or release.
Sound design uses this mechanism deliberately. A film composer doesn't add music at random—she identifies the emotional intent of a scene and selects a palette that pushes the viewer toward that state. A tense negotiation might pair low strings (threat, gravity) with sparse high notes (fragility). A montage of triumph pairs rising dynamics (crescendo) with major harmonies and rhythmic momentum (forward drive). The audience doesn't consciously parse "that's a minor ninth chord"; they feel the effect.
AI generation tools accelerate this workflow. Instead of waiting weeks for a composer to write a cue, you can describe the emotional intent ("tense, vulnerable, building to resolve") and generate a bed in minutes. Instead of licensing expensive stock music, you can remix and regenerate. The catch: your description must be precise. An AI model trained on thousands of film scores and game soundtracks has learned the same associations your brain has. "Sad music" is vague—it could mean funeral dirge or melancholic reflection. A solo cello in a minor key, sparse and high register, no accompaniment narrows the space and increases the chance the output matches your intent.
The real skill isn't the tool—it's learning to frame what you want in acoustic and emotional terms, then critique the output against your scene's actual needs. Does the rhythm serve the pacing? Does the timbre (tone colour) match the visual palette? Does the emotional shape match the arc?
WEEK 4: FINISHING YOUR FILM
Take Away the Music and the Scariest Scene is Just People Walking
Sound is the invisible story. A scene without music is neutral. A scene with the wrong music is laughable. A scene with the right music transforms the viewer's entire emotional experience.
Learn the architecture of sound design — emotion encoding, layering, silence — and how to score your film scenes with music that moves an audience. You will use these tools to write sound briefs that Claude and your composer can work from.
{"rubric": "Award 1 point per clip for correct emotion name; 1 point per clip if both acoustic choices are named correctly and mechanistically linked to the emotion (e.g., 'high register because it feels bright' is better than 'just sounds sad'). Full: 6 points.", "common_misses": ["Naming the instrument instead of the acoustic property (e.g., 'violin' instead of 'bright, high-register tone')", "Confusing intensity with emotion (e.g., 'loud therefore happy' — actually: intensity is orthogonal to valence)", "Auditory pareidolia — hearing lyrics or narrative instead of analyzing the acoustic pattern"]}Generate A Bed
Pick a 30-second scene from a film, game, or advertisement you know. Describe the emotional intent in 2–3 sentences (what the viewer should feel, why). Then write a prompt for an AI music generation tool specifying: (1) the emotion, (2) three acoustic properties (instrumentation, harmony type, tempo/rhythm character), (3) any constraints (e.g., "no lyrics", "build from sparse to full", "loop-friendly").