Atmosphere Engineering 2

Emotional Mix Ratios and Assignment Sheets

The Difference Between a Poster and a Scene

The rain-ritual specification in Part 1 named solemn resolve as its sole target emotion, at 100 percent purity. A state in which all eight layers point toward one emotion suits a poster or an album cover, but it is insufficient for a scene because actual human feeling does not remain that pure. A mourner at a funeral feels hunger amid grief; an employee hearing of a promotion calculates next quarter's burden amid joy. Audiences and readers use the presence or absence of such impurities to distinguish a poster from a scene. A character who sustains only one kind of emotion is read not as a person but as a symbol.

Impurity creates credibility because of the audience's habitual judgments. A highly purified single emotion is a pattern repeatedly learned from advertising and propaganda, so viewers may classify it not as emotion but as an intention to persuade. A detail irrelevant or contrary to the target emotion, by contrast, appears to have no reason to be fabricated and can therefore become evidence of reality. Low jokes exchanged by mourners do not damage the grief of a funeral hall; they make the place feel more real by the same structure.

Part 2 therefore asks two questions. The first is a problem of mixing: in what proportions should several emotions be combined in one scene? The second is a problem of assignment: in which materials among the eight layers should those minority emotions be placed? To state the conclusion first, a mix ratio is the language of design and an assignment sheet is the language of execution. Most failures occur not in mixing but in assignment. A scene with a poor ratio may merely feel bland; a scene with poor assignment fails to transmit the mixture at all.

Repeated exposure to advertising and propaganda is one explanation for this effect, not its established sole cause. The central observation is that details revealing the complexity of actual feeling can increase the credibility of characters and scenes.

The Lineage of Writing Emotions as Numbers

Writing emotions as proportions is not a new invention. Psychology has pursued related work for nearly a century. Robert Plutchik arranged eight basic emotions—joy, trust, fear, surprise, sadness, disgust, anger, and anticipation—in a circle and added an intensity axis to each. Within one family, annoyance, anger, and rage could therefore be distinguished as intensity values. He also organized a blending formula in which adjacent emotions form compound ones: joy mixed with trust becomes love, while fear mixed with surprise becomes awe. For practice, this model supplies intensity before mixture. If annoyance, anger, and rage are not different emotions but different values of one emotion, the habit of specifying emotional kind and intensity separately follows from this model.

James Russell's circumplex model changed emotion from a list into coordinates. It placed pleasure and displeasure on the horizontal axis, arousal and relaxation on the vertical, and represented every emotion as a position on that plane. Relationships among differently named emotions can thus be read as coordinates. Tension and excitement, for example, share an arousal value while occupying opposite sides of the valence axis; Part 3 examines the implications. The PAD model was an environmental-psychology tool Russell developed earlier with Mehrabian to measure emotional responses that spaces evoke in people. Its point of departure is therefore the same as this series' subject: the quantification of atmosphere. One practical use of Russell's plane is the search for substitute emotions. When a target emotion feels clichéd, the designer can move one step to a neighboring emotion in the same quadrant.

Paul Ekman's FACS is the notation system for facial expression promised in Part 1. It decomposes the face into more than forty action units and records each unit's contraction intensity in five grades, A through E. One expression can thus be written as a list of units and intensities: inner-brow raise C, lip-corner pull B. In this system, a genuine smile and a ceremonial smile differ as lists because a smile involving only the unit that pulls the lip corners has a different code from one that also contracts the muscles around the eyes. The latter distinction, known as the Duchenne smile, is a representative demonstration that expression is not an impression but readable data.

Literature has the same lineage. In his 1919 essay on Hamlet, T. S. Eliot introduced the concept of the objective correlative: rather than state an emotion as an adjective, a work should present the set of objects, situations, and events that constitutes its formula. Translated into the terminology of this series, the instruction is not to write an adjective but to present an assignment sheet. Literary criticism from a century ago and contemporary audio specifications have arrived at the same conclusion.

Industrial Tools for Emotional Mixing

These notations did not remain in papers; they became production tools. Blend shapes in 3D animation can apply smile 0.7 and brow tension 0.2 to a facial model simultaneously, producing a face that smiles while retaining worry. The two-colored memory orbs introduced late in Inside Out are a popular version of the same idea: emotion is a mixture rather than a single color. On set, directions such as “play this scene as anger 80, hurt 20” are actually exchanged. Actor training includes “playing the opposite,” the technique of performing a second layer of feeling contrary to the surface emotion of a line. Webtoon and illustration production use comparable documents. Character sheets routinely include expression charts that draw basic emotions at several levels of intensity, and assistants and contractors receive facial directions for a panel by coordinates on those charts.

AI tools accept this notation as an input format. Image-generation prompts assign multipliers such as 1.2 or 0.8 after words to specify their strength. Speech-synthesis services accept emotional styles as slider values. The AI Director in Left 4 Dead calculates player state as a numerical “intensity” and adjusts enemy placement and music in real time. Games also implement the mixture directly in music systems. Vertical layering in adaptive music separates one composition into percussion, bass, and melodic tracks, then adds layers one by one as the player approaches danger. Combat tension does not jump from zero to one hundred; it is produced by a continuous change in ratios. Writing emotion as a number and delivering it to a machine is not a forecast. It is already an operating method in several industries.

The Trap in Ratios: Mud

Entering a mix ratio literally, however, leads to failure. Give an image model solemn resolve 0.95 and doubt 0.05 as prompt weights, and doubt does not appear somewhere in the frame at five-percent size. Instead, the entire image becomes five percent murkier. Weights act nonlinearly across a result, so the minority emotion fails to establish itself through contrast and merely lowers the majority emotion's purity. Human perception can fail in the same way. Spread doubt lightly across the entire frame and the viewer receives not two coexisting emotions but one blurred solemnity. Audio engineers call a mix in which the lower midrange of several instruments is not separated but smeared together “mud.” Failed emotional mixing has precisely this form.

In prose, mud usually appears as hedging adverbs. The king was perhaps desperate. The crowd somehow seemed not to believe. For some reason the sky looked indifferent. Spread doubt over every sentence in this manner and the paragraph as a whole merely loses conviction; doubt never establishes itself anywhere. Concentrate the same amount of doubt in one sentence describing one person's gaze, and the remaining sentences can preserve pure solemn resolve while doubt receives an exact coordinate.

Acting has the same failure mode. Tell an actor that a scene is sad and have them add a little sadness to every line, and the performance comes out pressed beneath a single layer of tone. This is why directors say not to spread an emotion over everything but to plant it at one point. The diagnostic for mud is simple. Look at the finished work and ask where the minority emotion is. If you can point to one place with a finger, it has been assigned. If you must gesture toward the whole frame, it is mud.

Blend shapes demonstrate the same phenomenon. Raise every channel a little in an attempt to enrich a face, and the result is not a face with overlapping emotions but a doll-like one in which no emotion can be read. Mud is not always an error, however. It may be used deliberately when confusion itself is the target, as in a scene where a character cannot identify their own feeling. Even then, the staging expresses one target—confusion—with high purity; it is not a failure to mix. Whether a state is intention or accident depends on a single question: did the creator specify it?

“The entire image becomes five percent murkier” is a rhetorical model used to explain mixing failure, not a measured value. Because weights operate nonlinearly in actual models, 0.05 must not be read as corresponding to an exact five-percent change in the image.

Beds and Objects

The structure of a solution already exists in audio. Dolby Atmos separates sound into bed channels, spread broadly through a space, and object channels, individual sound sources with specified positions. Rain can be laid down as a bed while a telephone ringing behind the protagonist is placed as an object. A distributed sound and a sound pinned to an exact coordinate can therefore coexist in one space while remaining distinct.

The rule of emotional assignment follows the same structure: assign the majority emotion to the bed and minority emotions to objects. Lay the 90 of solemn resolve across continuous materials that cover the frame—light, air, color, and ambience. Condense the 7 of doubt into discrete, positioned materials such as one prop, one gaze, or one movement. Scattering a minority emotion across broad channels creates mud; concentrating it in narrow channels creates contrast. The number seven percent is implemented not as screen area but as the number and scale of the materials assigned to it.

The eight layers have different aptitudes within this structure. Air, light, and color are continuous materials covering the entire frame and therefore suit the bed. A prop in the material layer, a gaze or movement in the body layer, one shot in the viewpoint layer, and a one-time sound are discrete materials with specified positions and moments and therefore suit objects. Time can serve both: overall rhythm as a bed, one stop or one-beat delay as an object. There is also a limit to the number of terms in a mixture. A scene has only a few available slots for discrete object materials, so one or two minority emotions are the practical maximum. Beyond three, the objects divide the audience's attention until none can be identified. The dominant principle from Part 1 applies to mixtures unchanged.

The owner of a minority emotion determines which object material to choose. If the emotion belongs to a character, assign it to gaze or movement in the body layer. If it belongs to the state of the world, assign it to a prop in the material layer. The doubt in the rain ritual is divided between the local official's gaze and a cracked ritual vessel. Suspicion belongs to a person and therefore goes to gaze in the body layer; evidence that the rite has decayed belongs to the world and therefore goes to a prop in the material layer. Even objects carrying the same emotion split into different layers according to their owners.

Speech can also receive an assignment, but placing a minority emotion directly in the content of a line easily creates mud or excess. Rather than give a doubtful character a line expressing suspicion, it is safer to delay the answer by one beat or let the character remain silent, without objecting, at the point where agreement is expected. The emotion is then assigned not to the line but to delay in the time layer and silence in the sound layer. A shared technique of prose and screenwriting is that dialogue carries information while emotion is assigned around its edges.

An object has temporal coordinates as well as spatial ones. An object first introduced at the decisive moment is easily judged contrived. Standard practice is to expose it once earlier in a neutral state and retrieve it when needed. Read Chekhov's principle that a gun hung onstage must be fired in reverse: a gun that will be fired must have been hung there beforehand. At object scale, synchronization—the fourth principle in Part 1—is implemented through this prior exposure. Exposure has an upper limit. The default is neutral exposure that merely confirms existence without attaching a function. Even if viewers cannot consciously recall the object, prior exposure has done its job when retrieval leaves a sense that they have seen it before. Hinting at function during exposure forecasts retrieval and becomes a spoiler.

Hitchcock's rule of size joins this principle. An object's size within the frame should equal its narrative importance at that moment. This is the basis for allowing an object normally buried in the background to occupy the foreground for one decisive shot. A key, a glass, or a cigarette can shift the center of gravity of a scene with a single close-up.

Assignment in Practice: Two Films

In the flooding of the semi-basement in Parasite, despair is the majority emotion laid across the bed. Rising black water, darkness, and rain occupy the entire frame. Amid them, Ki-jung sits on the lid of a backflowing toilet and lights a cigarette. The minority emotion—the resignation of someone accustomed to such conditions—is assigned to that single cigarette. From one shot, viewers receive even the family's history: this is not the first disaster they have endured. The cigarette is an ideal object for two reasons. Lighting it is a familiar everyday action and therefore grounds the resignation; at the same time, the small point of its ember creates a visual coordinate against the dark bed.

Yasujirō Ozu's Tokyo Story offers assignment in the opposite direction: reverse assignment. In scenes where the widowed daughter-in-law Noriko speaks about her dead husband and his parents, the majority emotion of grief is not spread anywhere across the frame. Lighting, composition, and sound retain their everyday values, while the whole of grief is assigned to one object—Setsuko Hara's smile. When a smiling face becomes grief's only carrier, the drop experienced by the audience is greater than it would be if the entire image were sad. What has long been called restraint in Ozu's direction is, from the perspective of an assignment sheet, a technique of concentrating the majority emotion in the minimum number of channels. This reverse assignment also operates over time. Near the end, when the grief compressed into one smile throughout the film emerges for the first time as words and tears before her father-in-law, the release of a long-maintained assignment creates the scale of the scene. Assignment is a spatial technique and also a temporal one: a technique of maintenance and release.

Where Does a Mix Ratio Come From?

The numbers in a mix ratio derive not from taste but from three sources. The first is the scene's narrative function. The work that the scene must perform in the story determines the majority emotion. Foreshadowing planted for the next phase determines whether a minority emotion exists and how high it may rise. A clue cannot be zero because it must exist, yet it can seldom exceed ten because it must not occupy the foreground. In this way, function produces a range for the ratio.

The second source is the viewpoint character. Even when the event is the same, a scene's mixture belongs to the person to whom the camera or narration is attached. Write the rain ritual from the king's viewpoint and solemn resolve becomes the majority; write it from the local official's viewpoint and doubt becomes the majority. Every assignment is reorganized accordingly. The third source is the chain of scenes. Doubt at 0.07 in one scene may become 0.15 in the next and 0.40 in the one after that. The curve traced by changing ratios is itself the story's progress. A scene's ratio is not an isolated value but a point on this curve. In working notation, a range such as 0.85 to 0.95 is safer than a fixed value such as 0.90. A mix ratio is a target, not a measurement, and decimal precision does not imply actual precision. The purpose of a number is not to display accuracy but to share priorities and upper limits.

Rain Ritual: A Mixing Specification

Rewrite the monochrome specification from Part 1 as a mixing specification and it becomes:

Scene: the first rain ritual, midday
Mix: solemn resolve 0.90 / doubt 0.07 / awe 0.03

Assignment
Solemn resolve (bed): dust and heat haze in the air; direct light without shadows; a yellow-brown monochrome palette; ambience close to silence
Doubt (objects): one local official in the crowd looks not at the sky but toward the upstream weir gate; one cracked ritual vessel on the altar
Awe (object): one low-angle shot looking up at the altar

The bed inherits nearly all the values from Part 1. Doubt is condensed into one gaze and one prop; awe into one camera angle. Doubt at 0.07 does not mean that doubt occupies seven percent of the screen. It means that, among the eight layers, only one point each in body and material is given over to doubt. The grounds for the numbers come from the three criteria in the previous section. Solemn resolve at 0.90 follows because the scene's function is to establish the weight of the rite. Doubt at 0.07 must exist because it foreshadows the next phase—the politics of water—but must not enter the foreground. Awe at 0.03 is a constant of the setting: this is a world in which the Mandate of Heaven is real. The provisional 95-to-5 mixture in Part 1 is revised to 90, 7, and 3 by separating awe into its own term.

The assignment of the official's gaze to the water gate follows synchronization, the fourth principle in Part 1. This value derives not from a director's preference but from the setting: whoever controls the water can create a drought. When rendered in prose, an object's assignment is also implemented through its position in a paragraph. The final sentence of a paragraph is the prose equivalent of a close-up, so the official's gaze belongs in that position by default. Choosing the two or three layers to narrate, as described under selective rendering in Part 1, and choosing their positions in the prose are the two axes of prose assignment.

The Language of Design and the Language of Execution

Mix ratios and assignment sheets have different roles. Use the ratio for design and review; use the assignment sheet for production directions and prompts. Enter “doubt seven percent” into an image model and it produces the mud described above. Enter “everyone in the crowd looks at the sky; only one person in the back row looks toward the water gate at frame left,” and the model receives an executable result. The human role in collaboration with AI lies precisely in this translation. Designing an emotion as a ratio and translating it into assignments by layer remains work that no present model performs on the creator's behalf.

Parts of speech matter when an assignment sheet becomes a prompt. An object must be specified with a quantifier and a position: only one person, in the back row, toward frame left. These are the essential grammatical forms of assignment language. Without them, the model treats the direction as a bed.

Review runs this procedure in reverse. Read the completed scene or generated result, estimate its resulting ratio, and compare that estimate with the design value. If the reverse-engineered value of doubt is zero, the assigned object has disappeared or is too small. If it exceeds twenty, the object is oversized or has invaded the bed. The difference between designed and reconstructed values becomes the revision instruction itself. Assigning this review to a third party or an AI can reduce adaptation in the creator's own eye. The author already knows where the object was placed and therefore tends to read the minority emotion as larger than it actually appears.

In collaboration, it is practical to place both languages together on one scene card. A five-line document—a scene number and one-line function, one line for the mix, and roughly three lines for assignments—gives the director, artist, contractor, and AI the same instruction. A single document format thereby secures the reduction in output variance described in Part 1.

Training: Reverse-Engineering Great Scenes

The exercise for developing this sense is to reverse-engineer memorable scenes. Choose one, write its mix ratio, and make a table showing which material in which layer carries each emotion. The procedure has five steps: watch the scene again and name two or three emotions you feel; estimate their proportions; locate on screen the materials that produce each one; classify those materials as beds or objects; and record the whole as one line of mix plus three lines of assignments.

Reverse-engineer ten scenes and two facts become visible. First, most memorable scenes are mixtures rather than single colors. Second, minority emotions are condensed into objects with almost no exceptions. Choose only scenes by one director and that director's assignment habits emerge as a list. That list of habits is the preset discussed in Part 3.

The concern that reverse engineering ruins appreciation has matters backward. The first viewing is as an audience member; reverse engineering belongs to the second viewing, as a maker. It is closer to seeing the same scene twice from two perspectives. An efficient set combines three recent works in one's own genre with two works in other media. Half the exercise lies in confirming directly that the structure of assignment remains the same even when the medium changes.

The Remaining Problem: Translating Adjectives

We now have a procedure for designing one scene through mixing and assignment, but one problem at the starting point remains. Narrative function, viewpoint, and sequence can set the range of a ratio. Yet when a brief asks for an adjective such as ominous, wistful, or cozy, we still lack the correspondence table that translates it into a mixture and a set of layer values.

Experienced creators carry these correspondences as tacit knowledge, but tacit knowledge cannot be delivered to collaborators or AI. An explicit dictionary is required. Part 3 is that translation dictionary. It expands frequently used adjectives into mix ratios and assignment sheets, and assembles numerical scales already used in industry alongside director-specific presets. To preview the first entry: tension and excitement are physiologically the same state of arousal. What separates them is not arousal but another variable.