Generate a 3-second ultra-realistic 4K video using start and end frames.
Single continuous handheld push-in.
Begin on her partially zoomed face.
The camera smoothly pushes toward her mouth as it opens wider naturally.
Realistic lip mechanics, natural moisture highlights.
End on an extreme close-up inside the open mouth.
No distortion.
No exaggerated anatomy.
Preserve realism and texture accuracy.
A new experiment with Seedance 2.0 on @mitte_ai: FACS.
For this video, I only used FACS codes in the prompt. I didn't describe the facial expressions in plain language at all.
FACS (Facial Action Coding System) is a system for describing facial expressions using individual muscle movements called Action Units (AUs), instead of general emotion labels like "happy" or "sad." It breaks the face into controllable components such as brow movement, eyelid tension, lip movement and cheek activation.
Even though it didn't follow all 14 Action Units perfectly, it still interpreted most of them surprisingly well.
Seedance 2.0 Prompt for the first 15s:
Use the provided character @[image1] as the fixed identity reference.
15s, 1:1, 14 beats, beat-synced, cinematic tight close-up, subtle neutral background, high facial clarity, slow micro push-in, shallow depth of field.
1: AU10
2: AU20
3: AU22
4: AU23
5: AU27
6: AU28
7: AU45
8: AU53
9: AU61
10: AU62
11: AU64
12: AU85
13: AU84
14: AU46
Uneasy, hypnotic, controlled mood. No monster transformation, no gore, no comedy, no text overlay, no watermark.
SHOT1
Close-up on A (female).
Eyes glossy. Lips trembling but held tight. A:“If you walk out, don’t come back pretending it was complicated.”
SHOT2
Close-up on B (male).
He avoids eye contact. Throat tight. Breath uneven. B:“I’m not pretending. I’m scared.”
SHOT3
Extreme close-up on A’s eyes.
Tear finally falls. Voice steady despite it. A:“Then stay scared. Just stay.”
SHOT4
Wide shot, both framed in doorway, rain spilling light behind them.
B drops his bag. Silence holds.💬 B:“Okay.”