OmniHuman Prompt Guide
This guide helps you craft high-quality prompts for OmniHuman—covering character setup, motion control, expressions, camera style, and advanced creative tips.
Core Principles
Treat prompt writing as storytelling. Use coherent and natural language instead of stacking isolated keywords.
Only describe dynamic events. Do not describe static visual details already visible in the input image (e.g., clothing, jewelry).
Follow clarity, non-contradiction, and minimal negation. Use specific and concrete descriptions and guide actions step by step.
Best Practice Template
Recommended Structure
Camera movement
+
Emotion of the speaking character
+
Speaking state
(talking / crying / singing / …)
(talking / crying / singing / …)
+
Specific actions
+
(Optional) Background events / Other characters' actions
Complete Example
"The camera slowly moves from the side to a medium-front shot.
A young woman is sitting by the window, calm, gently picking up the coffee cup and smiling as she talks to the camera.
A boy beside her quietly watches her and occasionally turns to the camera and smiles.
The lights in the background are gently flickering."
Note: Including "the character talks/sings" improves lip-sync performance.
Writing Examples
Continuous Actions & Movement Control
Use "first… then…" to choreograph multi-step actions.
Always include "talking" or "singing" if the character must speak.
"The man walks forward first, then stops and puts his hands on his hips while speaking.
Then he turns around to look at the explosion behind him, showing his back.
The clothes on his back have been blown to pieces."
"The camera follows. The man turns around, looks at the camera and walks forward, singing.
He first touches his collar with both hands, then spreads his hands and raises his head."
Speak First Then Gesture / Gesture First Then Speak
"The man quickly took the cigarette out of his mouth, then started speaking to the camera."
"The camera zoomed in. The woman spoke to the camera, and after finishing, turned around and ran backward."
Multi-Character Action Control
Describe interaction between characters.
Subject Detection
Required to specify who speaks or sings.
One subject → pass its
mask_urlMultiple subjects → pass list of
mask_urls
"Following the camera, the boy quickened his pace to catch up with the girl and then started talking to her.
The girl stopped, turned around impatiently, and stared at the boy."
"Following the camera, the man holding a torch on the left looked at the others while speaking.
Then he turned back to look into the camera."
Camera Movement Control
Precisely control movement type, speed, and end position.
Use "rapidly / quickly" for stronger responsiveness.
"Pan the camera to the right in a circular motion and focus on the man's front face.
The man quickly takes off his glasses and speaks while facing forward."
"Pan the camera to the right in a circular motion.
The character speaks while facing forward.
Then quickly zoom in and focus on the character's front face.
The character puts down the binoculars and speaks."
Emotional Level & Micro-Expression Control
"First, the man looked calm and spoke to the camera.
Then he suddenly became nervous and panicked, frowning tightly."
"First, the woman rolled her eyes disdainfully and spoke with contempt.
Then she supported her chin with her index finger.
Finally, she played with the box on the table with a proud expression."
Special Effects / Adding Unreal Elements
"Distorted air ripples appear.
The father and son are sucked into the depths like dust and disappear instantly,
leaving only the remote control and phone on the empty sofa."
"The chick puts on sunglasses, holds two guns, and speaks evilly."
Camera Focus Control (Depth of Field)
"The reporter reports to the camera.
The king behind collapses.
The camera focuses on the collapsed king.
The reporter quickly turns around and speaks while looking behind."
"The woman turns around and speaks while looking at the soldier behind.
The camera focuses on the soldier.
Then she turns back to speak to the camera."
Environmental Interaction Description
"A gentle breeze blows.
The character's clothes and hair flutter slightly.
The leaves in the background sway."
"The orangutan maintains a selfie pose and talks to the camera.
The volcanic lava behind gradually turns blue."
Precautions & Recommendations
Recommended Practices
Upload high-resolution input images.
Complex scenes may require multiple generations.
If prompts do not work well, try omitting them.
Avoiding Common Issues
Structural stability is best within ~15 seconds.
Very small face region may cause mouth-not-moving issues.
Characters leaving and re-entering the frame may lose consistency.
Extreme poses or angles may downgrade results.