A LOT OF PEOPLE WHO DO AID VIDEOS FOCUS ON THEIR FACES: FIVE OFFICIALS ARE REAL, THEIR SKIN IS DELICATE, THEIR EYES ARE NATURAL, AND THEIR SHADOWS ARE CINEMATIC. BUT WHEN THE VIDEO CAME OUT, IT WAS EASY TO SEE IT WAS AI。
The reason is simple: video is not static. It's enough to look good for a moment, but the video needs the character to stay real for a few seconds. As long as people stand too steady, their faces remain unchanged and their movements are not justified, the picture will appear rigid。
Real people don't always park there like statues. People look at their mobile phones while waiting for a car, they turn cups at a coffee shop, they look at the floors in an elevator and they get windy when they walk. It's these subtle moves and life reactions that make people seem to be going through a real moment。
Let's talk about thisAI VideoThe core skill of character tips: not just what the person looks like, but what the person is doing, why, and how the movement continues。
I. WHY IS IT EASY TO FAKE AN AI VIDEO CHARACTER
Look at this。
The characters, costumes, scenes are fine, but if you make a video of it, just write:
An ancient woman stood in her room with a letter in her hand, slowly looking at the camera, and the sleeves moved gently。
IT'S DYNAMIC, BUT IT'S LIKE AI. 'CAUSE WHY DOES A PERSON HAVE A LETTER? WHY DOES HE LOOK AT THE CAMERA? IS SHE WAITING FOR THE NEWS OR JUST READING SOMETHING IMPORTANT? IF THERE'S NO REASON, IT'S JUST A SHOW。
IT'S NOT JUST HOW PEOPLE MOVE, IT'S WHY PEOPLE MOVE。
For example:
The ancient woman stood in a quiet room with her hands and a letter she had just received. She used to look down at the letter, like reading half of a sudden stop, with her sub-finger feeling tight on the paper, and then slowly looking forward, with her expression turning from calm to mild anxiety. The cuffs move slowly, the camera moves, the picture is quiet and restrained, and there is an ancient film feeling。
In this way, the person is not simply “standing with the letter”, but is going through a moment when she reads something, so she stops, squeezes the paper, heads up, changes in emotions。
SO, THE FOCUS OF THE AI VIDEO ALERT IS NOT JUST A SIMPLE DESCRIPTION OF THE DYNAMICS, BUT A CLEAR STATEMENT:
Where the characters are, what they're doing, why they move like this, and how they react。
In the same picture, static looks at the beauty of the picture; when it's made on video, it's about whether the person is actually in a situation or not。
II. The heart of the video message: action is motivated by behaviour
A lot of people write video alerts that directly describe the action:
Women up。
Girl back。
Boys drink coffee。
People walk slowly。
Hair floats with the wind。
They're not wrong, but they're just "face of action." If there's no reason to move, it's easy for the video to become a setup。
Case
And look at this。
The person in the picture was standing in the ancient wind room with a fan in his hand, and the clothes, lights and scenes were complete. The static is already pretty good, but if it's a video, just write:
The woman stands in the room, holds a fan, smiles slowly, shakes softly, and the camera moves slowly。
This hint is not without movement. People laugh, fans move, cameras push. But the question is: Why did she shake the fan? Why did she look forward? Was she waiting for someone, or did she just hear something outside the door
If there's no reason, it's easy to act。
IT'S NOT JUST WHAT SHE DID, BUT WHY IT HAPPENED。
For example:
The ancient woman stood in a quiet library with a fan in her hand, like someone waiting outside the door. She would have looked down at the scrolls on the table and slowly looked forward after hearing the light footsteps from outside the door. The fan in his hand had moved softly and then slowly stopped, and his expression had moved from calm to an expectation of restraint. It's a little twirl, it's a little twirl, it's a slow move, it's an ancient film feeling, it's quiet and natural。
In this way, the person is not simply “stand with a fan”, but is in a clear position: she is waiting for someone。
So she looks up, she stops the fan, and her face changes。
It's also a “hand-in-hand fan”, which is used in a normal way to make a person move; a better way to write is to allow a person to move naturally because he hears a voice, waits for someone, sees a change。
When writing video alerts, you can ask more:
What triggered this move
People shake fan because they wait in peace。
The character stopped because he heard voices outside the door。
The character looked up because someone was close。
The person's face changed because of emotional disruptions。
When the action is motivated, the character in the video is no longer a pose, but a real moment。
III. Single-person video alert formula
When a single person video is made, many people write only one action, such as "head down" "eye up" "turning" and "doing", but the person that is written in this way tends to be like a swing。
A better way to do that would be to break the phrase down into a complete process。
The single-person video can be applied to this formula:
Personal identity + specific scene + current state + action cause + continuous action + microresponse + lens language
We'll use this as a case。
The image is of an ancient woman, a night garden, a lantern on which she sits alone in front of the scene of the crime on the porch, with a slight forwarding of her body and her eyes, and an entire person has a quiet and low feeling. It's a good picture for a video like "Persons are Disgusting"。
If only:
An ancient woman sits in the courtyard, and her head is down, the camera is moving slowly, and she feels like an ancient film。
This is a picture, but it's still space. Because it only wrote the results, not the writing process. Why do people think? Did she just read something or wait for something? Did she react with her head down? None of this was explained。
Open it according to the formula, and it becomes clearer:
Personal status: an ancient woman
Specific scene: Lights bright under the courtyard at night
Current state: sitting alone in front of a crime
Reason for movement: It's like I'm thinking of something, and I'm feeling a little low
Sequencing: lower hair, soft fingers to the table, then slow eyes to the side
Slight reaction: paused eyes, light breathing, thin hair, quiet and restraining
Scene language: slow advance, light view deep, night wind
, which can be written in:
An ancient woman sits alone under the porch of the night, with a light on her forehead. It was like she was thinking about something in her head, quietly down, with her fingers down at the table, and her body leaning forward. After a moment of pause, her eyes slowly moved to the side, with a slight loss in her mind. The character has a natural blinking and a mild breath, a twirl and a little restraint. The camera is moving slowly, it's shallow, it's ancient, it's quiet。
The advantage of writing in this way is that the person does not have one “low head” action, but a complete process of change:
Let's go
And stop
I'm not sure what I'm talking about
Finally bring out emotions。
This is the most important part of the personal video alert:
Do not write a single action, write about the person's current state, movement changes and micro-responses。
The same picture
It'll be empty if it's just a "women's heads down."
The video would be more natural if it were clear: "Why does she go out of her mind, how she moves and how she moves?"。
IV. Making people look like real people: add a weak life response
This is an ancient woman standing in the palace, dressed in fine clothes, sanctified scenes, who is not fit to do too much。
If only:
The ancient women stood in the palace, slowly turning their heads, their sleeves drifted, the lens pushed, and the ancient film sense。
This paragraph, although dynamic, remains easy to fake. Because the character is just following the hint and there's no real physical reaction。
In this sacrificial scene, the person needs no big action, but rather a light life response to make it real。
for example:
She stood still in the palace, waiting for a call. Her eyes would have looked to the side, slid down after a moment of pause, and her fingers were slightly tightened in her sleeve. The candlelights were smooth, the hair was swayed, and the sleeves were softly up and down as they were breathing. She's not acting clearly, she's restraining, she's holding her heart down. The camera is moving slowly, it's quiet, it's ancient。
There's not much action in this hint, but it's more like real people。
Because real people don't even stand still. She'll breathe, her eyes will stop, her fingers will be fine, and she'll react lightly to emotional changes。
This chapter highlights:
Realism in the video does not necessarily come from big moves, but from small reactions。
In particular, the more restrained the figures, the more detailed they will be:
The eyes are paused
Tighten your fingers
The respiratory cuffs rise and fall
Slight shake of hair
Keep your lips down
There's a slight change in the weight of the body
The expression goes from calm to restraint
These details do not destroy the image's sense of dignity, but they make people look like statues。
SO WHEN YOU WRITE AN AI VIDEO MESSAGE, YOU CAN ADD THE LAST SENTENCE:
The character is not completely static, with natural blinking, slight breathing, short pauses in the eyes, small movements of the finger, and overall restraint in the movement of nature and not exaggerating performances。
Even if he was just standing there, he would be more like a real person in a situation。
V. Multi-person video: one person moves, another reacts
Multi-person video is the easiest to fake because many people allow everyone to act at the same time。
For example, this map
If only:
The ancient women and men stood in the room and looked up slowly, the men approached her, the camera moved slowly, and the ancient film sense。
This paragraph, although interactive, is more superficial. Because why are two people close? Why are they looking? Are women nervous, hesitant, or are they holding back? If it's not clear, it's easy to frame。
The key to multi-person video is not to move as much as possible, but to act on one side and react on the other。
May be replaced by:
In the glare of the wind, women were in red dress and stood in front of men. The man whispered a word, and his body was slightly near her. The woman did not reply immediately, but looked at him, with a brief pause in her eyes, and her lips squeezed, as if she were squeezing. The man stopped in front of her and did not continue to approach. The camera is slowly advancing, the light is deep, the picture is quiet and tense, the ancient film sense。
It's like this
Men come in and talk
The women heard a pause, looked up and reacted with restraint
Men stop again, and a quiet pull。
It's more like a real video than "two people against each other."。
A multi-person tip can remember one thing:
Don't let everybody do it, let one trigger an event, another react。
for example:
One man comes near, the other one steps back。
One person speaks, the other silent for a moment。
One man reached out, the other one looked away。
One stopped and the other noticed emotional change。
One looked at the other, the other looked away and slowly looked back。
This type of interaction creates a real relationship between the characters rather than a stand-in。
So, instead of just saying, "What are they doing" when you write a multi-person video, you have to be clear:
Whoever acts first, who reacts, how light it is and what changes have taken place in the atmosphere。
wind up
AS AN AI VIDEO CHARACTER, DON'T JUST THINK ABOUT GETTING PEOPLE MOVING。
The fact that people look back, blink, walk doesn't mean the picture is real. What's really important is that there's no reason, no change, no micro-responses。
A person with a letter can just stand and swing; he or she can read something important, pause, squeeze paper, and look slowly。
The difference is:
The former is acting, the latter is going through things。
SO WHEN YOU WRITE AN AI VIDEO MESSAGE, REMEMBER ONE SENTENCE:
It's not a simple description of dynamics, it's a way for people to move because of something。