Transposing Across Media
Given a source entity, extract its essence (not its surface form), then re-express that essence using the native grammar of the target medium.
Example: Text → Image
Source (text): "A quiet rebellion — she stopped asking for permission."
Essence extracted: tension between restraint and defiance; quiet intensity
Target (image) brief:
- Composition: figure centered, still posture, but one small defiant detail (untied shoelace, raised chin)
- Lighting: low-key, single light source — restraint
- Color: muted palette with one saturated accent — the "rebellion"
- Negative space: large, to convey quiet
Progress:
- Step 1: Identify the source entity's medium-native features (what makes it work as text/music/idea)
- Step 2: Strip to essence — theme, emotion, structure, rhythm, tension/resolution, key relationships
- Step 3: Identify the target medium's native vocabulary (what elements that medium uses to carry meaning)
- Step 4: Map essence components onto target vocabulary — find structural analogues, not literal equivalents
- Step 5: Produce target artifact or detailed brief/spec for it
- Step 6: Validate — does the target artifact evoke the same essence independently, without referencing the source?
Step 2 in detail — essence extraction categories
- Emotional arc: what feeling state begins/ends, and the shape of the transition
- Structural logic: repetition, escalation, contrast, symmetry, surprise
- Core relationships: tension vs harmony, figure vs ground, individual vs group
- Rhythm/pacing: fast/slow, dense/sparse, regular/irregular
- Central image or motif: the one thing that, if removed, destroys recognizability
Step 3 in detail — native vocabularies by medium
| Medium | Native vocabulary |
|---|---|
| Text | word choice, sentence rhythm, metaphor, narrative structure |
| Image | composition, color, light, framing, focal point |
| Music | melody, harmony, tempo, dynamics, timbre, silence |
| Script/screenplay | dialogue, scene structure, action beats, subtext |
| 3D model | form, proportion, surface texture, negative space, silhouette |
| Dance/choreography | movement quality, spatial pathways, tempo, body relationships |
Example 1: Input: Music → Script. Source: a jazz piece that starts sparse (solo piano, hesitant), builds to a dense polyrhythmic ensemble section, then resolves to a single sustained note. Essence: isolation → chaotic convergence → resolved solitude (different from the start — changed by the journey) Output (script beat structure):
SCENE 1: Character alone, minimal dialogue, long pauses (sparse/hesitant)
SCENE 2: Rapid intercutting of multiple characters/plotlines, overlapping dialogue, rising stakes (dense/polyrhythmic)
SCENE 3: Character alone again, but changed — one held shot, no dialogue, different body language than Scene 1 (resolved, but not identical)
Example 2: Input: Idea → 3D model. Source idea: "modular trust — a system that gets stronger as more independent parts connect, but still works if one is removed." Essence: independence + interconnection + resilience through redundancy Output (3D model brief):
Form: cluster of distinct geometric units (not identical — implies independence)
Connection: visible physical links between units (rods, joints) — interconnection made literal
Structural test: design such that removing any one unit leaves the rest standing — resilience embodied in the physics of the model, not just implied
Material: consistent material across units (shared "trust") but varied scale/orientation (independence)
Example 3: Input: Text → Music. Source: a short story about a slow-building argument that ends in silence, not resolution. Essence: escalating friction, abrupt cutoff, unresolved tension Output (music brief): dissonant intervals that increase in frequency and volume, irregular meter mimicking interruption, ending on an unresolved chord (e.g., suspended 4th) cut off abruptly rather than allowed to decay.
- Translate function, not form. A plot twist might become a key change, not a literal "twist" visual.
- Always name the essence explicitly before touching the target medium — skipping this leads to literal, shallow translations.
- Use the target medium's limitations creatively — e.g., music can't show text, so meaning must live in sound alone.
- When multiple essence components compete, prioritize the one most load-bearing (ask: "if I cut this, does the source still work?").
- Validate blind: show the target artifact to someone unfamiliar with the source and check if the intended feeling lands.
- Literal transcription: converting surface details (a red dress in text becomes a literally red object in the image) while missing the underlying emotional logic.
- Medium-blindness: producing a brief that ignores what the target medium actually does well (e.g., asking a 3D model to convey wordplay).
- Over-compression: trying to carry every detail of the source into the target, producing a cluttered, essence-less result. Prioritize ruthlessly.
- Skipping validation: assuming the mapping worked without checking if the target stands on its own.