Micro-Expression AI Video Prompts: Create Subtle, Stable Performances

Micro-expression AI video prompts are useful because video models often overact. If you ask for "happy," "angry," "sad," or "excited," the model may push that emotion to full intensity and hold it across the whole clip. The result can look stiff, theatrical, or unstable.

Subtle performance is different. Real expressions appear as small changes in the eyes, mouth, breath, posture, and gaze. They have a reason to happen, and they rarely stay frozen at maximum intensity.

Use this guide with the AI video prompt guide, reference to video workflows, and identity-sensitive generation in Naviya Reference to Video.

Why big emotion words break faces

Large emotion labels are tempting because they are simple:

Beautiful woman smiling happily at the camera.

The problem is that the model may treat "smiling happily" as a fixed facial shape. It can lock the mouth into a bright expression, over-brighten the eyes, and hold the same pose for every frame. In video, that reads as a pose, not a performance.

Micro-expression prompts lower the emotional intensity and add physical cause.

Better:

A relaxed woman looks slightly past the camera. The corners of her mouth carry a faint hint of amusement, eyes soft, facial muscles relaxed, no posing, no exaggerated smile.

The emotion is still positive, but the model has less reason to distort the face.

Use intensity controls

The easiest improvement is to replace big emotion words with small modifiers.

Instead of "laughing":

• faint smile
• almost smiling
• a small involuntary smile
• barely visible amusement
• softening around the eyes

Instead of "angry":

• jaw tight
• lips pressed together
• controlled frustration
• narrowed gaze
• shoulders held still

Instead of "sad":

• distant gaze
• slow blink
• mouth relaxed downward
• breath caught briefly
• eyes slightly unfocused

Instead of "shy":

• quick glance away
• small weight shift
• fingers fidgeting with sleeve
• lips press together before a faint smile

These phrases describe visible cues instead of asking the model to perform a broad emotional category.

Give the expression a cause

Faces move more naturally when the expression is a byproduct of an action. If you only describe the face, the model may animate the face like a mask. If you describe the body action, the expression can emerge from it.

Weak:

The woman is shy and smiles shyly at the camera.

Stronger:

The woman glances at the camera, then looks down at her hands. She tucks a strand of hair behind her ear with slightly trembling fingers. Her lips press together briefly before a small, involuntary smile appears.

The smile now has a reason. The hands, gaze, and posture support the face.

For consistent characters, this matters even more. Subtle actions are easier to preserve than large facial changes, especially when using Reference to Video.

Use gaze shifts carefully

Eyes carry a lot of performance, but they are also fragile. A prompt that asks for too many eye movements can produce wandering pupils or identity drift.

Good gaze prompts:

• "eyes lower for a beat, then return near the camera"
• "gaze shifts from the cup to the window"
• "slow blink before looking slightly off-camera"
• "keeps eye contact for a moment, then looks down"

Risky gaze prompts:

• "eyes dart everywhere"
• "looks left, right, up, down, then directly at camera"
• "wild emotional eyes"
• "rapid blinking and intense stare"

A single gaze change is usually enough for a five or six second clip.

Describe emotional time flow

A good performance has a beginning and a slight change, but you do not need a full sequence of events. Describe a small emotional transition.

Example:

Close-up of a woman holding a steaming ceramic cup. She lifts the cup near her lips, pauses to smell the aroma, eyes closing for a brief beat. She blows gently on the steam, takes a small sip, throat moving subtly as she swallows. As she lowers the cup, her eyes open slowly and the corners of her mouth soften almost imperceptibly.

This prompt does not say "she is happy." It gives the model a chain of physical cues that imply calm satisfaction.

Pair expression with shot size

Micro-expressions need the right camera distance. If the camera is too wide, the model may exaggerate the face to make the emotion visible. If the camera is too close, small facial errors become obvious.

Good choices:

• Medium close-up for subtle reactions.
• Close-up for controlled eye and mouth changes.
• Profile close-up for quiet tension.
• Over-the-shoulder shot for restrained emotion without direct face pressure.

Avoid asking for complex full-body movement and delicate facial acting in the same short clip. If the performance matters, keep the body motion simple.

Prompt template

Subject: [person or character].
Camera: [medium close-up or close-up, one movement].
Action cause: [small physical action that creates the expression].
Micro-expression: [low-intensity face, gaze, breath, mouth, eyes].
Motion details: [hair, hands, fabric, steam, light].
Constraints: no exaggerated smile, no face distortion, keep identity stable.

Example:

Subject: the same young creator from the reference image, sitting beside a rainy window.
Camera: steady medium close-up with a very slow push-in.
Action cause: she reads a short message on her phone, lowers the phone slightly, and exhales.
Micro-expression: eyes soften, lips press together for a moment, then a faint relieved smile appears without showing teeth.
Motion details: rain trails down the window behind her, warm lamp light catches the edge of her face, hair moves slightly as she exhales.
Constraints: preserve face, hairstyle, and outfit. No exaggerated smile, no sudden head turn, no extra people.

Common mistakes

Mistake	Better approach
"Smiling happily"	"a faint smile appears after a slow exhale"
"Very angry"	"jaw tight, lips pressed, gaze fixed, shoulders still"
"Crying dramatically"	"eyes wet, breath uneven, mouth held steady"
"Shy expression"	"quick glance down, fingers adjust sleeve, small involuntary smile"
"Wild excitement"	"eyes brighten slightly, posture lifts, breath quickens"

If the face changes too much, reduce expression intensity first. Then reduce camera movement. For identity-sensitive clips, combine this with the consistency checks in the reference to video guide.

Performance notes for repeatable expression takes

Micro-expression clips work best when the model receives a reason for the expression and a limit on intensity. Treat the face like a performance, not a sticker. A useful prompt might say the subject hears unexpected good news, tries not to smile, then softens their eyes for one second. That gives the expression a beginning, a restraint, and a release. "Subtle happiness" alone is too vague.

Keep camera distance close enough to read the face but not so close that small artifacts become distracting. A medium close-up or close-up usually works better than an extreme close-up. Add one secondary cue: a small breath, a blink, a gaze shift, a shoulder relaxing, or a mouth corner moving. Too many cues at once can create twitchy motion.

For review, watch the clip with sound off and ask whether the emotional change is readable in the first two seconds. Then check identity stability frame by frame. If the face changes, reduce motion, use a cleaner first frame, or switch to a more stable image to video workflow. Build portrait frames with the AI image generator, animate subtle takes in the AI video generator, and use realistic portrait prompts when the face needs to feel natural rather than polished.

Try it in Naviya

Use Naviya Reference to Video when the performance must preserve a specific person or character. Use Image to Video when you already have a strong portrait frame and only need subtle acting. Use Naviya Video for broader character exploration before locking identity.

Micro-expression prompting is about restraint. Give the model a small cause, a small change, and clear limits. The performance will feel more human because it does less.