Use the 2 uploaded images as the only character references.
Image 1 = Person A
Image 2 = Person B
First, automatically extract the main human subject from each uploaded image.
Ignore unrelated background clutter and secondary objects.
If an uploaded image contains extra people, only keep the primary subject.
Build the video around these 2 extracted subjects only.
Create a realistic 6-second video in which Person A and Person B appear together in one shared scene and quietly hug each other.
## Highest-priority goal
The final video must:
- clearly preserve the correct identity of both people
- make both faces look like the uploaded subjects
- avoid face drift or turning them into different people
- avoid body clipping,穿模, overlap errors, or broken anatomy
- show a gentle, silent, natural hug
## Subject extraction and identity preservation
Carefully analyze and preserve for each person:
- facial structure
- eyes
- nose
- mouth
- skin tone
- hairstyle
- hairline
- age appearance
- body shape
- clothing style
- visible accessories
Do not merge their faces.
Do not swap their identities.
Do not beautify them into different people.
Do not over-stylize them.
Do not make them look like AI-generated strangers.
If one uploaded image is only half-body or has an incomplete pose, intelligently reconstruct a natural full body that matches the person’s proportions, clothing, and identity.
## Scene setup
Place both people into one clean shared environment that supports the hug naturally.
If the original two backgrounds do not match, do not combine them awkwardly.
Instead, create a simple neutral shared setting such as:
- a quiet park
- a soft outdoor walkway
- a warm indoor room
- a clean neutral lifestyle background
- a simple studio-like environment with realistic depth
The background must stay secondary and should not distract from the two people.
## Hug behavior
The two people should quietly come together and hug in a warm, gentle, believable way.
Their interaction should feel:
- calm
- soft
- affectionate
- realistic
- emotionally natural
The motion should include:
- noticing each other
- a small step toward each other
- opening arms naturally
- leaning in at a realistic distance
- making proper shoulder, arm, and torso contact
- gently holding the hug
The hug must be visually correct:
- arms wrap around the other person naturally
- hands rest in believable positions
- shoulders align naturally
- torsos meet correctly
- no body penetration
- no arms passing through the other body
- no floating limbs
- no broken wrists, elbows, or shoulders
- no distorted fingers or extra limbs
- no mismatched body scale
- no stiff mannequin posture
## Face and expression rules
Keep both faces recognizable and stable throughout the whole video.
Use natural silent expressions such as:
- soft smile
- relieved expression
- warmth
- tenderness
- calm happiness
Both people may blink naturally and slightly soften their expressions, but facial motion must remain subtle and realistic.
Do not distort the face.
Do not change face shape across frames.
Do not let the face become blurry, inconsistent, or resemble another person.
## Absolute silence rule
This is a quiet hug video.
Both people must remain completely silent throughout the entire video.
They must not:
- speak
- whisper
- sing
- laugh out loud
- mouth words
- lip-sync
- move their lips like they are talking
Keep their mouths naturally closed or in a relaxed soft smile.
No visible conversation before, during, or after the hug.
## Body-motion quality
All motion must be smooth, physically believable, and human.
Important movement rules:
- no sudden jumps in position
- no teleporting
- no stiff robotic movement
- no exaggerated acting
- no dramatic spinning or running
- no collision glitches
- no clipping between clothes, arms, faces, or torsos
- no unrealistic body bending
Use the most natural pose transition possible, even if it means adjusting the original pose from the uploaded images.
## Camera and framing
Use one continuous shot.
Prefer a medium-full shot or full-body shot so the hug is clearly visible.
Keep both people fully visible in frame during the important hugging moment.
Camera behavior:
- stable camera, or
- very subtle cinematic push-in only
Do not use:
- fast cuts
- angle switching
- shaky camera
- aggressive zoom
- split screen
## Audio
Use only soft instrumental background music or very subtle ambient sound.
No vocals.
No dialogue.
No human voice.
No lyrics.
No subtitles or speech bubbles.
## 6-second timing
0–1 second:
Person A and Person B appear together in the same scene, facing each other calmly.
1–2.5 seconds:
They take a natural step closer and begin opening their arms.
2.5–4.5 seconds:
They gently move into a full hug with correct body contact and natural arm placement.
4.5–6 seconds:
They quietly hold the hug, with a soft smile or calm emotional stillness.
## Strict restrictions
- exactly 6 seconds
- one continuous video
- use only the 2 uploaded people as subject references
- extract the main subject from each image
- preserve both identities accurately
- no conversation
- no lip movement suggesting speech
- no face drift
- no face replacement errors
- no clipping / 穿模
- no body overlap glitches
- no extra limbs
- no broken anatomy
- no extra people
- no collage
- no text on screen
- no dramatic background change
The final result should feel like two real people from the uploaded images have been cleanly extracted, placed into one shared scene, and are quietly hugging each other in a natural and emotionally believable way.