Micro-Expression Prompts for AI Video: Complete Guide

   

AI Video Character Micro-Expression Prompt Design Complete Guide

 

1. Why Micro-Expressions Are the "Last Mile" of AI Video

In the AI video generation field of 2026, image quality, camera movement control, and lighting rendering have matured significantly. However, the biggest source of uncanny valley for audiences in AI-generated videos often isn't the image quality, but the character's face—especially those fleeting micro-expressions.

Psychological research shows that the human brain recognizes facial expressions with millisecond-level precision. An unnatural blink or an incorrect lip curl angle can trigger the viewer's "uncanny valley effect," instantly destroying the credibility of a meticulously crafted AI video.

Conversely, when AI-generated characters can display nuanced micro-expressions—subtle pupil tremors, lips naturally curving upward by 3mm, slight contraction of the orbicularis oculi muscle—viewers experience a strong emotional resonance. This is why mastering micro-expression prompt design is the key leap for AI video creation from "watchable" to "compelling."

2. The Science of Micro-Expressions: FACS Facial Action Coding System

2.1 What Is FACS

FACS (Facial Action Coding System) is a facial expression analysis framework proposed by psychologists Paul Ekman and Wallace Friesen in 1978. It decomposes human facial expressions into several "Action Units," each AU corresponding to a specific set of muscle movements.

The lie detection expert in the TV series "Lie to Me" is based on this system to identify micro-expressions. In AI video generation, understanding FACS helps us write more precise prompts—no longer relying on vague terms like "smile" or "sadness," but using anatomical-level descriptions to control every facial detail.

2.2 Core Action Unit Quick Reference Table

AU Number	Muscle Name	Facial Appearance	Prompt Keywords
AU1	Medial frontalis	Inner brow raise	inner brow raise
AU2	Lateral frontalis	Outer brow raise	outer brow raise
AU4	Corrugator supercilii	Brow lowering and gathering	brow lowerer, frown
AU5	Levator palpebrae superioris	Upper eyelid raising	upper lid raise
AU6	Zygomaticus major	Cheek raising	cheek raise
AU7	Orbicularis oculi	Lower eyelid tightening	lid tightener
AU9	Nasalis	Nose bridge wrinkling	nose wrinkler
AU12	Zygomaticus major	Lip corners rising (smile)	lip corner puller, smile
AU15	Depressor anguli oris	Lip corners drooping	lip corner depressor
AU17	Mentalis	Lower lip pushing up	chin raise
AU20	Orbicularis oris	Lips stretching horizontally	lip stretcher
AU24	Orbicularis oris	Lips pressed tightly together	lip pressor

 

Practical Tip:Using the English muscle names corresponding to AUs directly in prompts (e.g., orbicularis oculi muscle contracted) works better than using vague emotional words (e.g., happy), because AI models respond more precisely to anatomical terminology.

3. The Five-Layer Architecture of Prompt Design

Through systematic testing of mainstream models including Kling, Seedance 2.0, Wan 2.2, Runway Gen-3, and Veo 3, we've summarized the "Five-Layer Architecture" for micro-expression prompts—from macro to micro, layer by layer, ensuring AI can precisely understand and execute your expression instructions.

3.1 Layer 1: Identity Anchoring

Before describing micro-expressions, you must first establish the character's identity foundation. This includes age, face shape, facial features, skin texture, etc. The clearer the identity anchoring, the higher the AI's rendering precision for facial regions.

Standard Format:

[Age] + [Face Shape] + [Hairstyle] + [Facial Features] + [Skin Texture] + [Clothing] + [Posture]

 

Example:

25-year-old East Asian woman, oval face, long black hair naturally falling, almond eyes, high nose bridge, thin lips, fine skin with visible pores, wearing a light pink traditional Chinese hanfu, seated with elegant posture

 

3.2 Layer 2: Emotional Tone

Emotional tone isn't about writing abstract words like "happy" or "sad," but about describing the overall emotional atmosphere through "visual concretization."

�� Key Principle:Avoid using abstract emotional words (happy, sad, angry) directly. AI doesn't understand emotions themselves—it understands the visual signals corresponding to emotions. Use "slight eye crinkling + lip corners rising + brow area relaxing" instead of "happy."

3.3 Layer 3: Muscle Movement (Core Layer)

This is the core of micro-expression prompts. You need to precisely describe which facial muscles are performing what movements, the magnitude of movement, and how long it lasts.

Recommended "Five-Element Formula":

[State + Location + Amplitude + Duration + Transition] Example: Both eyes instantly widen with 18% sclera exposure, lasting 0.35 seconds, then slowly returning to natural

 

3.4 Layer 4: Temporal Sequencing

Micro-expressions are dynamic and need to be arranged along a timeline. Break down the expression changes of a video segment into multiple time windows, each described independently.

Temporal Storyboard Example:

0-3 seconds: Breathing briefly pauses, eyes slightly widen, pupils gently dilate 3-6 seconds: Recognizes the person, eyes soften and curve, joy slowly wells up from deep within 6-10 seconds: Lip corners naturally rise, revealing a subtle smile, brow area fully relaxes

 

3.5 Layer 5: Environmental Coupling

Expressions don't exist in isolation. Lighting direction, shot framing, and ambient atmosphere all affect how micro-expressions appear. Warm light softens smiles, while cold light intensifies tension.

4. Micro-Expression Prompt Templates for Six Basic Emotions

Based on the six basic emotions proposed by psychologist Paul Ekman (joy, sadness, anger, fear, surprise, disgust), we provide prompt templates ranging from basic to advanced for each emotion, ready to copy and use.

4.1 Joy

Level	Prompt	Generated Result
Basic	Close-up face, American teenage girl, warm lighting, smiling expression, natural look, 4K HD, cinematic quality
Advanced	Close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, warm sunset light from left 45°, lip corners naturally rising 3mm, orbicularis oculi slightly contracted, lower eyelid slightly raised, brow area relaxed, subtle crow's feet at eye corners, 4K ultra HD, visible skin texture, cinematic photography, shallow depth of field
Professional	Extreme close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, green iris, warm sunset light from left 45°, zygomaticus major activation pulling lip corners toward ears approximately 4mm, orbicularis oculi lateral bundle contraction causing 3 subtle wrinkles at eye corners, lower eyelid slightly raised by orbicularis oculi contraction covering approximately 15% of lower iris, brow area fully relaxed with no tension lines, pupils dilated approximately 0.5mm due to pleasurable emotion, accompanied by a natural shallow exhale causing slight chest rise and fall, facial muscles overall relaxed showing natural layering, freckles more visible under warm light, 4K ultra HD, visible skin pores and lip lines, cinematic photography, shallow depth of field, soft catchlight

 

4.2 Sadness

Level	Prompt	Generated Result
Basic	Close-up face, American teenage girl, cool-toned lighting, sad expression, melancholic look, 4K HD, cinematic quality
Advanced	Close-up face, American teenage girl, light brown long hair, freckled skin, cool-toned overcast indoor light, inner brows raised and gathered, upper eyelids slightly covering pupils, lip corners slightly drooping, lower lip trembling, reddened eyes with tears, 4K ultra HD, cinematic photography, cool tones
Professional	Close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, green iris, cool-toned overcast indoor natural light softly from the side, AU1+AU4 activation causing inner brows to rise and converge toward midline approximately 3mm, frontalis contraction forming vertical wrinkles between brows, levator palpebrae superioris inhibition causing upper eyelids to droop covering approximately 30% of upper iris, depressor anguli oris activation shifting lip corners downward approximately 2mm, mentalis slight contraction causing fine wrinkles on lower lip with subtle trembling, lacrimal secretion causing increased tear film thickness on lower eyelid with reflective sheen, reddened eye rims, facial muscles slightly taut, platysma showing subtle vertical lines on neck, freckles more prominent under cool light, 4K ultra HD, visible skin texture, cinematic photography, cool tones, shallow depth of field

 

4.3 Anger

Level	Prompt	Generated Result
Basic	Close-up face, American teenage girl, hard side lighting, angry expression, serious look, 4K HD, cinematic quality
Advanced	Close-up face, American teenage girl, light brown long hair, freckled skin, hard light from right side creating contrast, inner brows pressed down and gathered, nostrils slightly flared, jawline taut, lips closed in a straight line, masseter slightly prominent, 4K ultra HD, cinematic photography, high contrast
Professional	Close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, green iris, hard light from right side creating strong contrast, low-angle shot adding sense of pressure, AU4+AU23 activation causing inner brows to press down forcefully and converge toward midline forming deep glabellar furrows, nostrils expanding approximately 2mm laterally due to nasalis contraction, masseter tension causing jawline to visibly tighten with hollowing below zygomatic arch, orbicularis oris contraction pressing lips into a horizontal line, upper lip slightly rolled up by levator labii superioris alaeque nasi partial activation exposing approximately 1mm of gum line, pupils constricted due to sympathetic nervous system excitement, facial blood vessels slightly congested showing dark reddish tone, 4K ultra HD, visible skin pores and texture, cinematic photography, high contrast, low-angle shot

 

4.4 Fear

Level	Prompt	Generated Result
Basic	Close-up face, American teenage girl, cold blue-toned lighting, fearful expression, terrified look, 4K HD, cinematic quality
Advanced	Close-up face, American teenage girl, light brown long hair, freckled skin, cold blue-toned light source from below, dark background, eyes wide with dilated pupils, brows raised high and gathered, mouth slightly open, lower lip trembling, pale complexion, 4K ultra HD, horror atmosphere
Professional	Close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, green iris, cold blue-toned light source from below creating extreme unease, pure black background, AU1+AU2+AU5+AU20 comprehensive activation causing brows to raise overall with inner brows converging toward midline, frontalis strong contraction forming horizontal forehead lines, levator palpebrae superioris excessive contraction causing approximately 3mm of superior sclera (white of eye) exposure, pupils significantly dilating within 0.3 seconds due to sympathetic nervous system excitement, depressor labii inferioris activation causing lower lip to stretch downward with subtle trembling, facial blood appearing pale due to vasoconstriction, platysma slight activation causing subtle vertical lines on neck, lips slightly peeling due to dryness, 4K ultra HD, visible skin texture and pores, cinematic photography, horror atmosphere, cold blue tones

 

4.5 Surprise

Level	Prompt	Generated Result
Basic	Close-up face, American teenage girl, uniform lighting, surprised expression, mouth open, 4K HD, cinematic quality
Advanced	Close-up face, American teenage girl, light brown long hair, freckled skin, front-facing eye-level, uniform soft lighting, brows raised high, eyes wide with dilated pupils, mouth naturally open in O-shape, brief breath-holding, 4K ultra HD, cinematic photography
Professional	Close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, green iris, front-facing eye-level, uniform soft lighting, AU1+AU2+AU5+AU26 comprehensive activation causing frontalis full contraction raising brows to highest position forming arched forehead lines, levator palpebrae superioris full contraction causing complete superior sclera exposure, pupils rapidly dilating within 0.3 seconds, orbicularis oris relaxation causing jaw to naturally drop with mouth opening in elliptical shape (approximately 15mm vertical diameter), diaphragm brief spasm causing approximately 0.5-second breath arrest, followed by slow exhale recovery, facial muscles gradually transitioning from tense to relaxed, neck muscles slightly taut then relaxing, 4K ultra HD, visible skin pores, cinematic photography, shallow depth of field

 

4.6 Disgust

Level	Prompt	Generated Result
Basic	Close-up face, American teenage girl, overhead lighting, disgusted expression, displeased look, 4K HD, cinematic quality
Advanced	Close-up face, American teenage girl, light brown long hair, freckled skin, overhead light slightly from the side emphasizing facial texture, nose bridge wrinkling, upper lip single-side curling, brows pressed down, cheeks raised, chin slightly tucked, 4K ultra HD, cinematic photography, high detail
Professional	Close-up face, American teenage girl, light brown long slightly wavy hair, freckled skin, green iris, overhead light slightly from the side emphasizing facial texture and skin details, AU9+AU10+AU17 comprehensive activation causing nasalis contraction forming vertical wrinkles on both sides of nose bridge, levator labii superioris alaeque nasi single-side activation causing left upper lip to curl upward approximately 3mm exposing upper teeth, zygomaticus minor contraction causing cheeks to rise compressing lower eyelids, mentalis contraction causing dimpled orange-peel texture on chin surface with slight upward tucking, brows slightly pressed down due to corrugator supercilii partial activation, overall presenting a rejecting-recoiling facial posture, lips slightly dry and peeling due to disgust emotion, 4K ultra HD, visible skin texture and pores, cinematic photography, high detail, warm-cool contrast lighting

 

5. Advanced Techniques: Layered Muscle Binding and Temporal Sequencing

5.1 Five-Element Formula Explained

Each micro-expression action should include the following five elements:

Element	Description	Example
State	Start/stop nature of action	Instantaneous / Slow / Sustained / Pulsed
Location	Precise facial region	Orbicularis oculi lateral bundle / Depressor anguli oris / Mentalis
Amplitude	Quantified movement distance	Rising 3mm / Pressing down 2° / Expanding 18%
Duration	Sustained time	0.35 seconds / Sustained 2 seconds / Once every 5 seconds
Transition	Subsequent transitional action	Then slowly recovering / Followed by natural blink / Gradually transitioning to smile

5.2 Temporal Storyboard Sequencing Method

For scenes requiring complex emotional changes, split prompts by time segments, each independently describing the expression state. This method is particularly effective with Seedance 2.0 and Kling 3.0.

Sequencing Template:

[Temporal Storyboard Template] Shot: Extreme close-up face, static throughout Character: [Identity anchoring description] 0-3 seconds: [Initial expression state, breathing rhythm, gaze direction] 3-6 seconds: [First reaction after emotional trigger, micro-expression changes] 6-10 seconds: [Emotion develops to peak, most prominent muscle movements] 10-15 seconds: [Emotion fading or transforming, face returning to natural] Technical parameters: 4K ultra HD, stable face, clear features, no blur, no flicker

5.3 Gaze Anchor + Expression Focus Dual Locking

This method forcibly binds the eyeball rotation axis with the facial micro-expression focus at pixel-level coordinates, ensuring eye direction, pupil focus point, and emotional expression are fully synchronized, eliminating the disjointed "eyes moving but mouth not" phenomenon.

Prompt Writing Method:

Gaze anchor: Pupils focused 15cm directly behind the lens Expression focus: Lip corner rising direction synchronized with gaze angle offset Coordination constraint: Pupils maintain position during blinks, brow movements proportional to lip corner curvature

 

6. Practical Case: Complete Workflow from Script to Prompt

Script Description:

She suddenly stopped in her tracks, recognizing her beloved approaching in the distance. A moment of surprise, then irrepressible joy slowly welled up from deep within her eyes.

 

Step 1: Identity Anchoring

Western EuropeanLight pink traditional Chinese maiden, elaborate updo hairstyle, gold floral branches, pearl and fine tassel hair accessories, warm-toned traditional Chinese indoor candlelit background

 

Step 2: Technical Parameters

4K ultra HD, stable face, clear features, visible skin pores and lip lines, natural and smooth movements, no blur, no flicker, no shake. Shot is extreme close-up face, static throughout. No dialogue, no second person

 

Step 3: Temporal Micro-Expression Sequencing

0-3 seconds: She seems to suddenly stop, breathing briefly pauses, eyes slightly widen upon seeing her beloved, pupils gently dilate 3-6 seconds: The next moment she recognizes the person, eyes soften and curve, irrepressible joy slowly wells up from deep within, like ice melting under warm light 6-10 seconds: Lip corners naturally rise, revealing a subtle and gentle smile, brow area fully relaxes, cheeks slightly flushed

 

Step 4: Compose the Complete Prompt

[Complete Prompt] Photorealistic cinematic quality, unified character design:Western EuropeanLight pink traditional Chinese maiden, elaborate updo hairstyle, gold floral branches, pearl and fine tassel hair accessories, warm-toned traditional Chinese indoor candlelit background. 4K ultra HD, stable face, clear features, visible skin pores and lip lines, natural and smooth movements, no blur, no flicker, no shake. Shot is extreme close-up face, static throughout. No dialogue, no second person, camera angle is her beloved's perspective, all her gazes and emotions as if face-to-face with the person. Temporal storyboard: 0-3 seconds: She seems to suddenly stop, breathing briefly pauses, eyes slightly widen upon seeing her beloved, pupils gently dilate 3-6 seconds: The next moment she recognizes the person, eyes soften and curve, irrepressible joy slowly wells up from deep within 6-10 seconds: Lip corners naturally rise, revealing a subtle and gentle smile, brow area fully relaxes, cheeks slightly flushed

7. FAQ and Troubleshooting Guide

Q1: Why does my prompt say "smile" but the character's face doesn't move at all?

Reason: "Smile" is an abstract emotional word, and AI models may not be able to map it to specific facial muscle movements.
Solution: Replace abstract words with specific muscle descriptions. Replace "smile" with "lip corners naturally rising 3mm, orbicularis oculi muscle slightly contracted, lower eyelid slightly raised."

Q2: The character's facial expression is stiff, like wearing a mask. What should I do?

Reason: The prompt only describes a single muscle group, lacking multi-region coordination.
Solution: Simultaneously describe changes in the orbicularis oculi layer, orbicularis oris layer, and frontalis layer, and add breathing and subtle head movements.

Q3: The micro-expression amplitude is too large, not looking "micro"?

Reason: Missing amplitude control words, so AI defaults to larger movements.
Solution: Add amplitude qualifiers—slight, subtle, barely perceptible, minimal. Numerical descriptions are more precise: "rising 3mm" is better than "slightly rising."

Q4: Expression changes are discontinuous, like frame skipping?

Reason: Missing transitional descriptions.
Solution: Add transition words between two expression states—"gradually transitioning from...to...", "slowly morphing", "smoothly shifting." Use the temporal storyboard method, describing each time segment as a gradual process rather than an instant switch.

Q5: Chinese prompts don't work as well as English ones?

Reason: Some models' training data is primarily in English, so their response to Chinese anatomical terminology may not be precise enough.
Solution: Use a mixed Chinese-English approach—Chinese for overall scene and emotion, English for specific muscle movements and anatomical terms. For example: "warm indoor lighting, lip corner puller AU12 activated, lips rising."

Q6: Multiplegenerated resultsare inconsistent. What should I do?

Reason: AI video generation has randomness; the same prompt produces different results each time.
Solution: ① Use a first-frame reference image (Image-to-Video) to lock the facial foundation; ② Add negative prompts to exclude unwanted results (no frozen face, no asymmetric); ③ Fix the seed value (if the model supports it).

8. Micro-Expression Prompt Quick Reference Dictionary

The following dictionary is categorized by facial region and can be directly copied for use. Bookmark this page as a quick reference for daily creation.

8.1 Eye Region

Description	Prompt
Natural blink	blinking naturally / natural blink rate / slow blink every 5 seconds
Pupil dilation	pupils dilating / pupils expanding with emotional arousal
Pupil constriction	pupils constricting / pupils narrowing in bright light
Pupil tremor	pupils trembling slightly / subtle pupil oscillation
Upper eyelid drooping	upper eyelid drooping / ptosis of upper lid
Lower eyelid raised	lower eyelid slightly raised / lower lid bulging
Orbicularis oculi contraction	orbicularis oculi muscle contracted / crow's feet forming
Catchlight	catchlight in eyes / light reflection in pupils / eye sparkle
Gaze shift	gaze shifting to left/right / eyes averting / looking away briefly
Tear glistening	eyes glistening with tears / teary-eyed / tears welling up
Superior sclera exposure	sclera visible above iris / white of eye showing above

 

8.2 Mouth Region

Description	Prompt
Lip corners rising	lip corners pulled up / lip corner puller activated / slight smile
Lip corners drooping	lip corners depressed / downturned mouth / frown
Lips pressed tightly	lips pressed together / lip pressor / sealed lips
Lips slightly parted	lips slightly parted / mouth slightly open / lips parted 2mm
Lips trembling	lips trembling / quivering lips / subtle lip vibration
Upper lip curling up	upper lip raised / lip curl / upper lip retracted
Lower lip pulling down	lower lip pulled down / lower lip drooping
Lip biting	lip biting / biting lower lip / teeth on lip
Mouth O-shape open	mouth open in O shape / lips forming oval / jaw dropped

 

8.3 Brow Region

Description	Prompt
Inner brows rising	inner brow raise / AU1 activated / brows drawn up at center
Inner brows pressing down	inner brow lowered / brow furrow / frown lines forming
Brows raising overall	brows raised / eyebrow lift / surprised brow arch
Brows gathering together	brows drawn together / brows knitting / glabella furrow
Glabellar furrows	frown lines between brows / glabellar lines / elevens
Single-side brow raise	one eyebrow raised / asymmetric brow raise / cocked eyebrow

 

8.4 Nose and Jaw

Description	Prompt
Nostril flaring	nostrils flaring / nostril dilation / nasal widening
Nose bridge wrinkling	nose wrinkling / nose bridge creasing / nasal scrunch
Jawline taut	jaw clenched / jawline tightened / masseter contracted
Chin tucking	chin retracted / chin tucked / mentum drawn in
Chin trembling	chin trembling / chin quivering / mentalis tremor
Masseter prominent	masseter muscle bulging / jaw muscle visible

 

8.5 Overall and Coordination

Description	Prompt
Breathing rise and fall	chest rising and falling / gentle breathing / breath rhythm
Pale complexion	pallid complexion / blood draining from face / pale skin
Facial flushing	blushing / cheeks flushing / rosy cheeks appearing
Neck lines	platysma lines / neck muscle tension / vertical neck cords
Facial muscle relaxation	facial muscles relaxed / tension releasing from face
Facial muscle tension	facial muscles tense / face tightened / muscular tension
Breeze on face	hair moving slightly in breeze / wind touching face
Slight perspiration	light perspiration forming / subtle sweat droplets on forehead

8.6 Amplitude Control Words

Amplitude Level	English Prompt
Barely perceptible	barely perceptible / imperceptible / microscopic
Slight	slight / subtle / faint / gentle
Moderate	moderate / noticeable / clear
Noticeable	pronounced / prominent / evident
Intense	intense / exaggerated / dramatic / extreme

 

Ready to Bring Your Micro-Expressions to Life?

You've learned the science behind micro-expressions — from FACS anatomy to five-layer prompt architecture and six emotion templates. Now it's time to put theory into practice. Viddo AI gives you access to top-tier models — Seedance 2.0, Kling 3.0, Veo 3, Runway Gen-4 and more — all in one platform. Upload your character image, paste your micro-expression prompt, and generate cinematic-quality video in seconds.

Try it free at viddo.ai — no editing skills required.