Story
Why your scene is boring even though nothing is wrong with it
The image is handsome, the music is right, and still nobody watches to the end. Usually because everything is where it started. Five questions that catch it.
By: Kovács-Réz BálintAI film directing, storytelling
This is the most common disappointment: the technical side worked, the look is professional, and people scroll past anyway. The image is rarely the problem. The problem is that the scene never happens.
A scene works when something is DIFFERENT at the end. It isn't the location that has to change. It is the character's position, knowledge or resolve.
The five questions
Before you generate anything, answer these five about the scene. One sentence each is enough. Whichever one you cannot answer is where the fault is.
- What do they want?What does the character want in these ten seconds? Not in life. Here, now.
- What is in the way?If nothing is, there is no scene, only presentation.
- What is at stake?What do they lose if it fails? With nothing at stake, nobody is rooting for them.
- What do they do?At the peak, do they ACT, or do things merely happen to them?
- What changes?How is the last second different from the first?
Short video needs a different shape
The classic three-act shape is too slow for a video of a minute or two: by the time the conflict is built, the viewer has long gone. In short form, three stretches work, each with a hard time limit.
| Stretch | Time | What it does |
|---|---|---|
| Hook | 0–10 s | a visual contradiction that reads without words |
| Escalation | 10–40 s | quicker cutting, camera moving in, rising stakes |
| Payoff | 40–60 s | a visible emotional change on the character's face |
The hook has the hardest job: it has to work with the sound off, because most people start watching muted. A visual paradox (something in the frame that doesn't add up) stops a scroll more reliably than any caption.
When the AI cannot do lip sync
Today's generators are weak at matching speech to mouth movement. That looks like a limitation, and in fact it leads straight back to one of the oldest tools in the language of film.
Cut the same expressionless face after two different images and the viewer supplies the emotion. A child, then an explosion: tragedy. The other way round: fireworks. Same face, two different scenes.
While lip sync is unreliable, put the conflict between the character and their SURROUNDINGS, and let the cut create the emotion.
One question that settles everything
A weak story answers „what happens". A good one answers „why does this matter to the person we are following". Get that into one sentence and the scene very nearly assembles itself, and the viewer starts to care as well.
Sources
- Wikipedia: Kuleshov effect
- Digiday (2016): 85 percent of Facebook video is watched without sound
Common questions
Why is my AI video boring when the picture looks great?
Usually the image is not the problem. The scene simply does not happen: at the end, everything stands exactly where it did at the start. A scene works when something is different at the end. It is not the location that has to change, but the character's situation, knowledge or resolve.
What questions should I ask before writing a scene?
Five questions filter out a weak scene: what does the character want, what stands in the way, what is at stake, what do they actively do at the peak, and what has changed by the last second. One sentence is enough for each. If you cannot answer one of them, that is where the problem is.
How should a one-minute short video be structured?
The classic three-act structure is too slow, so three phases with hard time limits work better: a hook from 0 to 10 seconds, escalation from 10 to 40, and resolution from 40 to 60. The hook has to work without sound, and the resolution should show a visible emotional change on the character's face.
What can I do if the AI cannot handle lip movement well?
Do not build the scene on speech. Put the conflict between the character and their surroundings, and create the emotion through editing: if you cut the same expressionless face in after different images, the viewer projects the emotion onto it. It is one of the oldest tools in film language.