Start a project
WorkBlogResourcesAboutContact

Blog/Anatomy of a shot

Why a 14-Second Fight Prompt Needs Every Beat on the Clock

My duel prompt was losing its opening four seconds to setup. Exact action beats gave each clash and the unfinished exit a place.

By Jeff Zambrano··6 min read

Why a 14-Second Fight Prompt Needs Every Beat on the Clock

The model was spending the first four seconds of a fourteen-second duel merely getting the action started.

My earlier prompt described what happened. It did not say when. The model distributed the available time and used the opening to start action that should already have been underway.

I saw the cost directly: the beginning of the shot was being used to start something that should already have been happening.

The rewrite needed a clock. It also needed a first frame that agreed with that clock.

I could not fix the opening just by labeling a paragraph “0 to 4 seconds.” If the men began standing still, the prompt would still contain a start that needed to happen.

The scene context therefore establishes the fight as underway. The first frame makes the instruction specific:

FIRST FRAME / BLOCKING
Both men are already at a full run, mid-stride, weight forward, blades up and travelling. @character-toshiro-mabe-formal enters frame left moving right; @character-ryo-sagi-brown enters frame right moving left. Sand is already in the air. Tight two-shot, cut off at mid-thigh.

There is action in that frame before the first exchange has finished. The men have direction and weight. Their blades are moving. The sand shows that their feet have already disturbed the ground.

The left and right entries also place the movement. I am asking two identified bodies to close toward each other, with the frame cutting them at mid-thigh.

That gives the opening something definite to show. The clock can then govern the exchange instead of carrying the entire burden of making the scene active.

The timing rule I borrowed called for a hook every 3 to 5 seconds and a turn every 5 to 7. For this duel, that becomes four beats across fourteen seconds.

I kept the timing idea because it addressed the problem in this shot. Other parts of the same framework called for changing settings and working vertically. Those did not fit the film. Its 21:9 frame and repeated setting were deliberate.

A framework can contribute a useful rule without deciding the film around it.

Here is the complete action block:

ACTION
0 to 4 seconds: both men cover the last four metres at running speed and the blades meet the first time, steel on steel, and slide apart.
4 to 8 seconds: the second exchange, closer, with the camera whipping across to the far man as his hips turn into the cut.
8 to 11 seconds: the third clash, the tightest of the three, the two guards locked for a fraction of a second at chest height.
11 to 14 seconds: they break and the next attack is already loading, and the shot ends with that movement still travelling.

The first interval includes an arrival, contact, and release. It gives the opening a destination within the fight.

Both men cover the remaining distance. Their blades meet. The blades slide apart. The interval therefore has more to accomplish than showing the swordsmen run toward each other.

The four metres belong to the blocking. The four seconds belong to the whole opening beat. The prompt places the first contact and its separation inside that same interval.

That distinction matters when I read it back. I am checking whether the words allocate the opening to an exchange, because the earlier version had allowed that time to become preparation.

The next interval links the camera and the body. The second exchange moves closer. The camera whips toward the far man as his hips turn into the cut.

The hip turn gives the camera a reason to move. The time range gives that coordinated event a place in the take.

Without that connection, I would have a camera instruction in one paragraph and a sword action in another, with less direction about their relationship. Here, the movement of the frame belongs to a particular exchange.

The third interval tightens the encounter again. The guards lock briefly at chest height.

This beat has its own image. It does not simply repeat the instruction to make the swords clash. The distance has closed, the guards meet, and the contact holds for a fraction of a second.

That brief lock also gives the following separation something to break away from. The action block carries the shot through contact and release.

The last interval is where I protect the exit. The men separate, but another attack is already loading. The shot ends while that movement continues.

I need that final instruction on the clock because an unfinished exit still occupies time. It cannot survive only as a general wish for the duel to feel ongoing.

The final seconds contain an action, just as the first seconds do. The beginning enters movement. The ending leaves movement.

The camera’s separate block supports that design. It stays at chest height, 1.5 to 4 metres from the bodies. It moves with the fight, whips laterally, and arrives a fraction late to the clashes.

Those directions describe a camera that catches an encounter in pieces. The timings organize those pieces without turning them into separate shots.

The format block is explicit:

FORMAT MODE
One continuous shot, the camera holds the take to the last frame. Fast lateral whips inside the shot are camera movement.

The time boundaries divide the instructions for reading. The camera holds one take through them.

That leaves the action block free to state what changes within the shot: the first contact, the closer second exchange, the tight third clash, and the break into another attack.

I also need the supporting physical behavior to keep running across those boundaries. The prompt gives sand weight, lets the robes lag behind the torso, and asks the blades to carry their swings through. A timestamp does not replace those instructions.

Nor does the written schedule prove that a generated take obeyed it. I was fixing the opening delay. The rewrite makes the intended allocation explicit. A claim about the resulting footage would need footage to support it.

What the clock gives me at the prompt stage is a way to inspect the request. I can see whether a clash has a time range, whether the camera movement belongs to it, and whether the ending still has room to travel.

Today, put exact time ranges beside the action already in your prompt, then check that the first frame is active enough to begin the opening beat immediately.


You just watched one shot come apart, decision by decision. None of them landed there by luck.

ShareXLinkedInWhatsApp
Jeff Zambrano

Jeff Zambrano is a Madrid-based AI film director. His debut, Borrowed Wounds, reached the Top 50 of the Higgsfield Global Film Festival out of 6,700 films. He writes these notes from his own productions.

WhatsApp