# Skill: Animation & Motion Techniques for Sequential AI Image Generation

## What
A comprehensive reference for applying professional animation and comic techniques to frame-by-frame AI image generation. Makes sequential still images feel like motion.

## Why
AI image generation produces isolated stills. Without deliberate motion language, sequential panels feel like a slideshow of unrelated photographs. These techniques — drawn from Disney, Pixar, anime, and comics — give the model clear physical instructions that create perceived motion across frames.

---

## 1. Disney's 12 Principles — Applied to Sequential Stills

Each principle below is translated from "frames per second" thinking to "panel-to-panel" thinking.

### 1.1 Squash and Stretch
**Animation:** Objects deform — flatten on impact, elongate in motion. Gives weight and flexibility.
**In stills:** Describe the deformation state for the moment you're capturing.

| Moment | Prompt language |
|--------|----------------|
| Ball hitting ground | "basketball compressed flat against court floor, sides bulging outward" |
| Character jumping | "legs compressed in crouch, torso squashed low, arms loaded" |
| Character at apex of jump | "body elongated vertically, arms stretched overhead, legs trailing below" |
| Face reacting to shock | "eyes stretched wide, jaw dropped open, cheeks pulled taut" |

**Key:** Pick ONE phase of the squash/stretch cycle per panel. Don't describe the full motion — describe the freeze-frame.

### 1.2 Anticipation
**Animation:** A preparatory action before the main action. Pulling back before punching. Crouching before jumping.
**In stills:** The anticipation frame is often MORE dramatic than the action frame. It's the loaded spring.

| Action | Anticipation panel prompt |
|--------|--------------------------|
| Throwing a ball | "arm pulled far behind head, weight shifted to back foot, eyes locked on target, body coiled like a spring" |
| Opening a door | "hand gripping doorknob, shoulder tensed, weight leaning slightly back — about to pull" |
| Standing up from chair | "hands pressing down on armrests, torso leaning forward, feet planted flat, weight shifting forward" |

**Prompt pattern:** `[character] is [preparatory pose], [weight distribution], [tension indicator] — about to [action]`

**PRO/FLASH allocation:** Anticipation frames are KEY FRAMES. Use PRO model. The anticipation IS the storytelling.

### 1.3 Staging
**Animation:** Present an idea so it is unmistakably clear. One idea per shot.
**In stills:** Every panel has ONE primary read. The viewer should understand the action in under 1 second.

Rules:
- **Silhouette test:** If you blacked out the character, would you still understand the pose? Describe poses that read in silhouette: arms away from body, clear angles, no limbs hidden behind torso.
- **One action per panel.** If two things are happening, split into two panels.
- **Camera serves the action.** A fist hitting a face? Close-up. A character feeling small? Wide shot, tiny figure.

**Prompt pattern:** `[camera angle] capturing [the ONE thing this panel communicates]. [Character] is [clear pose with silhouette-readable limbs]. [Supporting elements subordinate to main action].`

### 1.4 Straight-Ahead vs Pose-to-Pose
**Animation:** Straight-ahead = draw frame 1, then 2, then 3 (spontaneous). Pose-to-pose = plan key poses first, fill gaps later (controlled).
**In sequential AI generation:** This maps DIRECTLY to our model allocation.

| Method | Animation | AI Sequential |
|--------|-----------|---------------|
| Pose-to-pose | Key poses planned first | PRO model on key frames (establishing shots, emotional peaks, action climaxes) |
| In-betweens | Junior animators fill gaps | FLASH model on transition frames (walking between locations, reaction shots) |
| Straight-ahead | Used for fluid, chaotic motion | Use when you want organic unpredictability — fire, water, crowd chaos |

**Planning rule:** Before writing prompts, mark each panel as KEY or TWEEN. Key frames get:
- Full character descriptions
- Precise camera angles
- PRO model allocation
- No ref_panel shortcuts — describe the full scene

Tween frames get:
- "Same scene" + delta description
- FLASH model
- Heavy ref_panel reliance
- Minimal prompt — describe only what CHANGED

### 1.5 Follow-Through and Overlapping Action
**Animation:** When a character stops, their hair, clothes, and loose parts keep moving. Different body parts stop at different times.
**In stills:** This is the #1 way to make a still image feel like it was captured mid-motion.

**Prompt elements for follow-through:**
- Hair: "hair still swinging to the left from the sudden stop"
- Clothing: "coat hem still billowing forward though she has stopped walking"
- Accessories: "pendant still swaying, catching light"
- Limbs: "arms still trailing behind from the sprint, not yet at rest"

**Prompt pattern:** `[character] has just [completed action]. [Primary body position]. But [secondary element] is still [continuing previous motion direction].`

**Example:** "Same scene. She has stopped at the edge of the cliff. Feet planted, body rigid, leaning back slightly. But her scarf is still streaming forward over the void, and her hair hasn't settled yet, blown forward by her own momentum."

### 1.6 Slow In and Slow Out (Ease)
**Animation:** More frames at the start and end of a motion, fewer in the middle. Objects accelerate and decelerate.
**In stills:** This controls HOW MANY panels you allocate to different phases of an action.

| Phase | Panel density | What to show |
|-------|--------------|--------------|
| Starting to move | HIGH (2-3 panels) | Subtle weight shifts, first tentative movement, building momentum |
| Mid-motion | LOW (1 panel or skip) | Blur, speed lines, peak velocity — one panel captures it |
| Coming to rest | HIGH (2-3 panels) | Deceleration, settling, follow-through, final pose |

**Example — character sitting down in a chair (4 panels):**
1. Slow in: "turning toward the chair, one hand reaching back to find the armrest"
2. Slow in: "lowering herself, knees bending, weight transferring, hand gripping armrest"
3. Mid: "dropping the last few inches into the seat, body compressing"
4. Slow out: "settled into the chair, adjusting position, clothes settling around her"

**For AI prompts:** Don't distribute panels evenly across an action. Cluster them at the transitions. The beginning and end are where storytelling lives.

### 1.7 Arcs
**Animation:** Natural motion follows curved paths, not straight lines. Arms swing in arcs. Heads turn in arcs. Thrown objects follow parabolas.
**In stills:** Describe the ARC position, not just the endpoint.

- Wrong: "his arm is raised above his head"
- Right: "his arm is sweeping upward in a wide arc, currently at the 2 o'clock position, hand trailing behind the elbow"

**Prompt pattern for arc motion:** `[body part] is [moving verb] in [arc shape], currently at [clock position / percentage / spatial marker], [leading part] ahead of [trailing part]`

### 1.8 Secondary Action
**Animation:** Actions that support the main action without taking focus. Whistling while walking. Drumming fingers while listening.
**In stills:** Layer one secondary action into EVERY panel. It adds life.

**Examples:**
- Main: character explaining something. Secondary: "absent-mindedly stirring coffee with one hand"
- Main: character running. Secondary: "backpack bouncing rhythmically, straps flapping"
- Main: character arguing. Secondary: "one foot tapping, free hand clenching and unclenching"

**Prompt structure:** Describe the main action first (2-3 sentences), then add: "Meanwhile, [secondary action that reveals character or adds physicality]."

### 1.9 Timing
**Animation:** The number of frames between poses determines speed. Fewer frames = faster.
**In stills:** Panel count IS timing. This is the most direct translation.

| Speed | Panel allocation | Prompt language |
|-------|-----------------|-----------------|
| Lightning fast | 1 panel for entire action | "mid-punch, fist a blur against his jaw, impact ripple visible" |
| Fast | 2 panels (wind-up + impact) | Panel 1: anticipation. Panel 2: result. |
| Normal | 3-4 panels | Beginning, middle, end, settle |
| Slow/dramatic | 5-7 panels | Moment-to-moment. Every subtle shift gets a panel. |
| Frozen moment | 1 panel held for multiple beats of narration | "time seems to stop — she stands in the doorway, hand still on the handle" |

**Rule:** If an action gets only 1 panel, describe it at its MOST DYNAMIC moment — the apex, the impact, the peak. Not the beginning or end.

### 1.10 Exaggeration
**Animation:** Push poses beyond realistic for emotional clarity. Bigger reactions, wilder takes, more extreme poses.
**In stills:** AI models tend toward neutral poses. You must PUSH in your prompts.

| Emotion | Realistic | Exaggerated (use this) |
|---------|-----------|----------------------|
| Surprise | Eyes slightly wide | "eyes impossibly wide, jaw dropped to chest level, body recoiling backward, arms flung out" |
| Anger | Furrowed brow | "veins visible on forehead, teeth bared, fists white-knuckled, leaning aggressively forward with shoulders raised to ears" |
| Joy | Smile | "beaming so hard her eyes are squeezed shut, arms thrown overhead, rising onto tiptoes" |
| Exhaustion | Slouching | "collapsed over the desk, one arm dangling toward floor, papers stuck to sweaty cheek" |

**Prompt rule:** Whatever emotion the panel needs, describe it at 130% intensity. The AI will render it at roughly 100%. If you prompt at 100%, it renders at 70% — flat, boring, lifeless.

### 1.11 Solid Drawing (Solid Construction)
**Animation:** Characters should feel three-dimensional, with weight, volume, and proper perspective.
**In stills:** Describe the 3D space the character occupies. Use the Blender Camera Mindset skill.

**Prompt elements for solidity:**
- Weight: "feet pressing into soft ground, slight impression visible"
- Volume: "broad shoulders filling the doorframe, head ducked to clear the lintel"
- Perspective: "seen from slightly below, foreshortened legs, torso receding upward"
- Lighting confirming form: "overhead light casting shadow under brow ridge, nose shadow falling across left cheek"

### 1.12 Appeal
**Animation:** Characters should be interesting to watch — clear design, readable shapes, charisma.
**In stills:** This is about CHARACTER DESIGN, not individual panels. Lock it in the object sheet.

Appeal elements for object sheets:
- Distinctive silhouette (could you recognize them as a shadow?)
- Color signature (one dominant color per character)
- Characteristic pose or gesture (how they stand when idle)
- Facial feature hierarchy (what reads first — eyes? mouth? brow?)

---

## 2. Key Frame vs In-Between — The PRO/FLASH Model

### The Pose-to-Pose Pipeline for AI Comics

**Step 1: Script → Mark Key Frames**
Read the script. Mark every moment that is:
- An establishing shot (new location, new scene)
- An emotional peak (the moment the character realizes, breaks, decides)
- An action climax (the punch lands, the door opens, the reveal)
- A scene transition (cutting to a new time/place)

These are KEY FRAMES. Everything between them is IN-BETWEEN.

**Step 2: Write Key Frame Prompts First**
Key frame prompts get full treatment:
- Complete character description (even if they appeared before)
- Full environment description
- Precise camera angle and lens
- Style token at full strength
- NO "Same scene" shortcuts — describe from scratch

**Step 3: Write In-Between Prompts**
In-between prompts describe only the delta from the previous panel:
- "Same scene."
- What moved, what changed
- ref_panels carries the visual DNA

**Step 4: Generate Key Frames First (PRO model)**
Generate ONLY the key frames. Review them. These are the anchors. If a key frame is wrong, everything chained from it will drift.

**Step 5: Generate In-Betweens (FLASH model)**
With solid key frames as ref_panels, generate the in-betweens. FLASH model is sufficient because the visual DNA is already established.

### Frame Type Decision Matrix

```
Is this the first panel of a new scene?     → KEY FRAME (PRO)
Does the character's emotion peak here?      → KEY FRAME (PRO)
Does a major action complete here?           → KEY FRAME (PRO)
Is this a dramatic reveal?                   → KEY FRAME (PRO)
Is this walking/transitioning between beats? → IN-BETWEEN (FLASH)
Is this a reaction to the previous panel?    → IN-BETWEEN (FLASH)
Is this showing time passing?                → IN-BETWEEN (FLASH)
```

---

## 3. Multiple Moving Objects — Clarity Techniques

### 3.1 Relative Position Anchoring

When 2-3 characters are in the same frame, describe their positions RELATIVE to each other and to fixed landmarks.

**Bad:** "Three soldiers are in the trench."
**Good:** "Three soldiers in a muddy trench. The sergeant stands at left, one boot on the fire step, binoculars raised. The medic kneels at center, bent over an open kit bag on the duckboard. The private huddles at right, pressed against the trench wall, rifle across his knees."

**Spatial anchoring formula:**
```
[Character A] is at [position], [pose], [interaction with environment].
[Character B] is at [position relative to A], [pose], [what they're doing].
[Character C] is at [position relative to A and B], [pose], [what they're doing].
```

### 3.2 Staggered Movement

Not everything moves at once. In real life, actions cascade. Apply this to sequential panels.

**Principle:** Between any two panels, only 1-2 elements should change significantly. Everything else holds or shifts subtly.

**Example — three characters reacting to an explosion (4 panels):**

| Panel | Character A (soldier) | Character B (medic) | Character C (private) |
|-------|----------------------|---------------------|----------------------|
| 1 | Ducks immediately, helmet flying off | Still upright, hasn't reacted yet | Still upright, hasn't reacted yet |
| 2 | On the ground, covering head | Flinching, turning toward blast | Beginning to duck, rifle slipping |
| 3 | Still down, looking up | Dropped to knees, shielding kit bag | Flat on ground, arms over head |
| 4 | Rising to one elbow, scanning | Already reaching for bandages | Still flat, trembling |

**Why this works:** Each character reacts on a different timeline. The stagger creates a wave of motion that the viewer reads as a single event unfolding through time.

**Prompt pattern for stagger:** In each panel prompt, explicitly state what each character is doing, including "still [previous pose]" for characters who haven't moved yet.

### 3.3 Focal Priority — Who Moves, Who Holds

In any multi-character panel, designate:
1. **Primary mover** — the character whose action drives the story. Describe them FIRST and in MOST DETAIL.
2. **Reactor(s)** — characters responding. Describe their reaction in 1 sentence.
3. **Holders** — characters who haven't changed. "Same position as previous panel" or simply don't mention them (ref_panel carries them).

**Example prompt:**
"Same scene. [PRIMARY] The sergeant lunges forward, grabbing the radio handset from the shelf, body twisted mid-reach. [REACTOR] The medic looks up from his kit, startled, hands frozen mid-bandage. [HOLDER] The private remains pressed against the wall, unchanged."

### 3.4 Differential Speed

When characters move at different speeds, use prompt language that encodes velocity:

| Speed | Prompt indicators |
|-------|-------------------|
| Static | "motionless", "frozen", "planted", "hasn't moved" |
| Slow | "slowly turning", "gradually shifting weight", "inching forward" |
| Normal | "walking", "reaching", "turning to face" |
| Fast | "lunging", "snapping around", "arm a blur of motion" |
| Explosive | "erupting from position", "blur of movement", "frozen mid-leap at impossible angle" |

Describe each character's speed explicitly when they're moving at different rates.

---

## 4. Action Lines and Motion Cues in Still Images

### 4.1 Physical Motion Indicators (No Drawn Lines Needed)

AI models don't reliably draw speed lines. Instead, describe the PHYSICAL EVIDENCE of motion:

| Technique | Prompt language | When to use |
|-----------|----------------|-------------|
| **Hair displacement** | "hair streaming behind her", "hair whipped forward across face" | Any fast movement |
| **Cloth behavior** | "coat billowing outward from the spin", "skirt pressed against legs by wind of movement" | Walking, running, turning |
| **Particle trails** | "dust cloud trailing behind", "water droplets spraying from wet hair" | Running, landing, shaking |
| **Lean angle** | "body tilted 30 degrees into the turn", "leaning hard into the sprint, torso nearly horizontal" | Running, turning, dodging |
| **Weight shift** | "all weight on front foot, back heel lifted", "center of gravity low and forward" | Any dynamic pose |
| **Motion blur (selective)** | "fist a blur of motion, rest of body sharp and crisp" | Fast punches, swings, throws |
| **Ground interaction** | "rubber streaks on floor from sudden stop", "cracks radiating from impact point" | Landings, stops, impacts |
| **Air displacement** | "loose papers on desk scattering from the rush of air", "candle flame bent sideways" | Fast movement past objects |

### 4.2 The Lean Angle Rule

The faster a character moves, the more they lean into the direction of motion.

```
Standing:    0 degrees (vertical)
Walking:     5-10 degrees forward lean
Jogging:     15-20 degrees
Sprinting:   25-35 degrees (nearly diagonal)
Lunging:     40-50 degrees (extreme)
```

Always specify the lean: "body angled 20 degrees forward into the run."

### 4.3 Implied Motion Through Environment

The environment can show motion the character can't:
- "curtains still swaying from where she pushed past them"
- "chair still spinning slowly where he shoved it aside"
- "puddle ripples expanding from where his foot just left"
- "door swinging open, hinges still vibrating"

These are EVIDENCE of action that just happened. Extremely effective in the panel AFTER the action.

---

## 5. Pixar/Disney Staging for Multi-Character Compositions

### 5.1 Focal Hierarchy

In any frame with 2+ characters, the viewer's eye must go to the RIGHT character first. Control this with:

| Technique | How to prompt |
|-----------|---------------|
| **Size** | Focal character larger (closer to camera or physically bigger) |
| **Position** | Focal character at a power position (rule of thirds intersection) |
| **Motion vs stillness** | Focal character is the only one moving; others are still |
| **Lighting** | Focal character is better lit; others in shadow or flat light |
| **Color** | Focal character wears the brightest/most saturated color |
| **Focus** | "shallow depth of field, [focal character] sharp, [others] slightly soft" |
| **Gaze direction** | Other characters LOOKING AT the focal character directs viewer there |

**Prompt pattern:** Describe the focal character first, with the most detail. Then describe others in decreasing detail. The prompt structure mirrors the visual hierarchy.

### 5.2 Character Relationship Through Composition

| Relationship | Composition | Prompt language |
|-------------|-------------|-----------------|
| Intimacy / agreement | Characters close together, leaning toward each other | "standing close, shoulders nearly touching, both leaning in" |
| Conflict | Characters facing each other with space between, leaning forward aggressively | "facing each other across the table, both leaning forward, 3 feet of hostile space between them" |
| Power imbalance | One character higher/larger, other lower/smaller | "the boss stands over her desk, looming; she sits small in her chair, looking up" |
| Isolation | Character separated by space, framing, or obstacles | "he stands alone at left frame, a pillar separating him from the group clustered at right" |
| Alliance | Characters aligned, facing the same direction | "standing shoulder to shoulder, both facing the horizon, arms crossed identically" |

### 5.3 The Triangle Composition

For 3 characters, arrange them in a triangle (not a line). One character at the apex, two at the base.

**Prompt:** "The three of them form a triangle — [Character A] stands closest to camera at center, [Character B] is behind and to the left, [Character C] behind and to the right. [Character A] is the clear focal point."

---

## 6. Anime Action Techniques

### 6.1 Impact Frames

In anime, the moment of impact is often shown as an abstract burst — pure energy, no background detail. The frame BEFORE shows the wind-up, the frame AFTER shows the aftermath. The impact itself is abstracted.

**For AI sequential generation:**

| Frame | What to generate |
|-------|-----------------|
| Pre-impact (KEY FRAME) | Full scene. Anticipation pose. Fist drawn back, target visible. Clear staging. |
| Impact | "Abstract burst of white and gold energy radiating from center point. No characters visible — pure kinetic force. Radial lines emanating from impact center." |
| Post-impact (KEY FRAME) | Full scene. Result visible. Target reeling. Attacker in follow-through pose. |

**When to use:** Dramatic moments only. A punch that matters to the story. Not every action — overuse kills the effect.

### 6.2 Speed Lines and Environmental Blur

Rather than asking the AI to draw literal speed lines (unreliable), describe the effect:
- "background is a horizontal blur of color — only [character] is sharp and in focus"
- "everything behind her streaks into parallel horizontal lines of motion"
- "radial blur emanating from the point of impact, sharp at center, dissolved at edges"

### 6.3 Smear Frames (Translated to Stills)

Smear frames in animation are single distorted frames showing a character stretched between two positions. In stills:
- "her arm is a blur of motion connecting two positions — drawn back and extended forward simultaneously, like a double exposure"
- "his head appears to be in two places at once, turning so fast it traces an arc of afterimages"

Use sparingly. One smear-style panel per major action sequence maximum.

### 6.4 Dramatic Hold

Anime often PAUSES on a character's face before or after an action. Static frame, dramatic lighting, maybe wind in hair. No motion — pure emotion.

This is a KEY FRAME. Full PRO model treatment:
"Close-up on [character]'s face. [Precise expression]. [Dramatic lighting — strong side light creating deep shadows]. Hair lifted by wind. Eyes [specific emotion]. No motion — a frozen moment of [what they've just realized/decided/felt]."

---

## 7. Scene Transitions — Scott McCloud's Taxonomy Applied to AI Generation

### 7.1 The Six Transition Types

| Transition | Time gap | Space gap | Panel density | AI prompt approach |
|-----------|----------|-----------|---------------|-------------------|
| **Moment-to-moment** | Seconds | None | Highest | "Same scene. [Tiny change]." ref_panels essential. FLASH model. |
| **Action-to-action** | Seconds-minutes | None-slight | Medium | "Same scene. [Action completed/advanced]." ref_panels. |
| **Subject-to-subject** | None | Same scene | Medium | Different camera angle, different character focus. May need new establishing context. |
| **Scene-to-scene** | Hours-years | Different location | KEY FRAME | Full new description. No ref_panels. PRO model. New establishing shot. |
| **Aspect-to-aspect** | None | Same location | Variable | Show different DETAILS of the same moment. Clock, hands, face, window. |
| **Non-sequitur** | N/A | N/A | N/A | Rarely used. Deliberate disorientation. |

### 7.2 Moment-to-Moment (Video Feel)

This is our bread and butter for video-feel comics. Each panel advances time by 1-3 seconds.

**Prompt pattern:**
```
Panel N: "Same scene. [Character] has [micro-action]. [One thing changed]. [Everything else unchanged]."
```

**Example — character picking up a phone:**
1. "Phone buzzing on desk, screen lit up. Her hand pauses mid-typing on laptop."
2. "Same scene. Her eyes have shifted to the phone. Hand still on laptop keys. Phone still buzzing."
3. "Same scene. She reaches for the phone, fingers closing around it, lifting it from desk."
4. "Same scene. Phone raised to eye level, screen illuminating her face. Laptop forgotten, hands off keys."

### 7.3 Action-to-Action (Standard Comics)

Skip the micro-moments. Show the key beats.

**Example — same phone pickup:**
1. "Phone buzzing on desk."
2. "Same scene. She's answering the phone, pressed to ear, expression shifting to concern."

Two panels instead of four. Faster pacing.

### 7.4 Scene-to-Scene (Location/Time Jump)

This is a KEY FRAME. No ref_panels. Full description.

**Transition prompts that establish the jump:**

| Jump type | Prompt approach |
|-----------|----------------|
| Time jump (same place) | Show the same environment with changed conditions: different lighting, season, decay/growth, different objects present |
| Location jump | Full establishing shot of new location. Wide angle. Character small in environment to establish scale. |
| Flashback | Shift color palette: desaturated, warmer, softer focus. "Memory-like quality, slightly hazy, warm sepia undertone." |
| Flash-forward | Colder palette, sharper edges, slightly overexposed. |

### 7.5 Aspect-to-Aspect (Mood Building)

Show multiple angles/details of a single moment. Slows time to almost nothing. Builds atmosphere.

**Example — tense meeting:**
1. "Close-up: a pen tapping rapidly on the conference table surface"
2. "Close-up: sweat bead rolling down the side of a neck, collar tight"
3. "Wide: the long conference table, everyone frozen, all eyes on the door"

No character moves between these panels. Time is essentially frozen. You're showing the FEELING of the moment from multiple angles. No ref_panels between them (different camera angles), but consistent environment description.

---

## 8. Prompt Templates for Common Motion Scenarios

### Walking Toward Camera (3-panel)
```
P1: "[Character] visible at far end of hallway, full body, walking toward camera. Hallway stretching behind them, overhead fluorescents. Arms at sides, mid-stride, left foot forward."

P2: "Same scene. [Character] closer now, visible from knees up. Left arm swinging forward, right arm back. Coat swaying with stride. Background hallway compressed by proximity."

P3: "Same scene. [Character] filling most of frame, waist-up. Close enough to see expression — [emotion]. Right hand reaching for the door handle at frame edge."
```

### Fight Sequence (4-panel)
```
P1 (KEY/Anticipation): "[Attacker] has pulled right fist far back, weight loaded on back foot, torso twisted away from [target]. [Target] is unaware, back partially turned, mid-sentence. Two feet apart."

P2 (Impact): "Extreme close-up of [attacker]'s fist connecting with [target]'s jaw. Moment of impact — cheek compressed, spit droplets suspended in air. Background is radial blur."

P3 (KEY/Follow-through): "[Target] staggering backward three feet from where they stood, head snapped to the side, arms flailing for balance. [Attacker] in follow-through pose — arm fully extended, weight shifted to front foot. Chair knocked sideways between them."

P4 (Aftermath): "[Target] collapsed against the far wall, hand on jaw, looking up. [Attacker] standing over them, fist still clenched at side, breathing hard. Overturned chair on floor between them."
```

### Character Sitting Down (Slow In/Slow Out, 4-panel)
```
P1: "[Character] standing beside the armchair, one hand on the armrest, beginning to turn."

P2: "Same scene. [Character] halfway down, knees bent at 90 degrees, one hand still on armrest for support. Weight transferring. Coat bunching at the waist."

P3: "Same scene. [Character] settling into the seat, back pressing against cushion. Hands releasing armrests. Coat settling around them."

P4: "Same scene. [Character] fully seated, one leg crossing over the other. Leaning back. Hands finding their resting position — one on knee, one on armrest. Settled."
```

### Two Characters Talking — Emotional Shift (Subject-to-Subject, 4-panel)
```
P1: "Medium two-shot. [A] at left, animated, leaning forward, gesturing with both hands. [B] at right, arms crossed, skeptical expression, leaning back slightly."

P2: "Same scene. Close-up on [A]'s face. Mid-sentence, eyebrows raised, slight desperation in the eyes. Mouth open, caught between words."

P3: "Same scene. Close-up on [B]'s face. Expression softening — arms uncrossing (visible at bottom of frame), brow relaxing. Something [A] said is landing."

P4: "Same scene. Return to two-shot. Postures reversed — [A] now leaning back, spent, hands in lap. [B] leaning forward, uncrossed arms, palms open on the table."
```

---

## 9. Common Mistakes

| Mistake | Why it fails | Fix |
|---------|-------------|-----|
| Prompting at 100% intensity | AI renders at ~70%. Result looks flat. | Prompt at 130%. Push poses, expressions, angles. |
| Everything moves at once | Looks chaotic. Eye doesn't know where to go. | Stagger: 1-2 elements change per panel. |
| Describing motion instead of position | "She is running" gives static result. | "She is mid-stride, left knee driving forward, right foot pushing off ground, torso at 25-degree lean." |
| Equal detail for all characters | No focal hierarchy. | Primary character: 3 sentences. Secondary: 1 sentence. Background: skip or 3 words. |
| Using speed lines in prompt | AI draws them inconsistently. | Use physical evidence: hair, cloth, particles, lean angle, blur language. |
| Skipping anticipation frames | Action feels like it came from nowhere. | Always pair major actions: anticipation panel + action panel minimum. |
| No follow-through | Characters feel like posed mannequins. | Add secondary motion: hair settling, clothes catching up, objects still swaying. |
| Same camera angle every panel | Feels like security footage. | Vary shot size (wide/medium/close) and angle per the action's needs. |
| Distributing panels evenly across action | Boring pacing. | Cluster panels at beginnings and endings (slow in/slow out). Skip or compress middles. |
| "The character looks angry" | Vague emotion. AI defaults to mild frown. | "Teeth bared, veins visible on forehead, fists white-knuckled, leaning forward with shoulders raised to ears." |

---

## 10. Quick Reference — Motion Language Cheat Sheet

### Verbs That Encode Speed
```
SLOW:    drifting, easing, settling, sinking, melting into
MEDIUM:  stepping, reaching, turning, shifting, adjusting
FAST:    snapping, lunging, whipping, slamming, erupting
INSTANT: frozen mid-[action], caught in the blur of, a streak of
```

### Body Part Motion Descriptors
```
HEAD:    snapped toward, slowly turning, tilted, ducked, thrown back
EYES:    darting to, locked on, widening, narrowing, tracking
ARMS:    sweeping upward in arc, trailing behind, windmilling, bracing
HANDS:   clenching, spreading wide, snatching, releasing, trembling
TORSO:   twisting toward, collapsing forward, arching back, coiling
LEGS:    planted wide, driving forward, buckling, crossing, kicking
FEET:    pivoting, stomping, sliding, lifting onto toes, digging in
```

### Environmental Motion Evidence
```
WIND:     hair streaming, clothes pressed flat, papers scattering
IMPACT:   cracks radiating, dust erupting, objects bouncing
HEAT:     air shimmering, sweat visible, fabric clinging
COLD:     breath visible, shoulders hunched, arms wrapped tight
WATER:    splashing outward, droplets suspended, ripples expanding
GRAVITY:  objects falling, hair floating upward (freefall), settling dust
```