Why Four Wounds Would Not Fit Into a 25-Second Fight

Four wounds did not fit into the 25-second fight.

The earlier prompt divided the shot into six blocks. That gave each block 4.1 seconds. The revision reduced the fight to three marks across four blocks, giving each block 6.25 seconds.

That was the point where an action could read.

I had been asking the shot to carry too many events, while also spending prompt space on instructions that either never reached the model or worked against the result. Reducing the wound count was part of a larger correction, but the timing exposed the central problem first.

The arithmetic made the cut

I wrote down the pacing decision exactly:

**3. Cuatro heridas en 25 segundos no caben.** Con seis bloques eran 4,1 segundos cada uno.
Bajado a tres marcas y cuatro bloques: 6,25 segundos, que es donde una acción se lee.

The original structure gave every block the same basic problem. It had 4.1 seconds to establish the movement, make the contact legible, and get into the next event.

The revised structure gave each block 6.25 seconds.

That extra time was not used to add another wound. It was the reason to remove one. The goal was to let the remaining actions register rather than pack the timeline with more incidents.

This is where my fight prompt changed from an inventory of things that happen into a paced sequence. Three marks were enough. Four were not.

The prompt was also being cut before the model saw it

The earlier version was roughly 9,000 characters. The web interface cuts the prompt at around 5,787 characters without warning.

The entire POSITIVE LOCKS block sat beyond that cutoff. Those were the instructions that mattered most, and they never reached the model.

This changed how I judged every sentence in the revision. The 25-second shot had a timing budget, but the prompt also had a space limit. Keeping more action blocks meant describing more transitions, more contacts, and more states. At the same time, the prompt was already long enough to lose its final section silently.

The failure did not appear as an error message. The prompt simply arrived incomplete.

Reducing the fight to four action blocks made the timing clearer. It also made it possible to place the essential instructions inside the part of the prompt that would actually be received.

Negative instructions were feeding the wrong words

The previous prompt contained sixteen negative orders. I had found that the model does not process the “no.” It sees the word.

I had written “slow motion” seven times while trying to prevent slow motion.

The revision replaced those negative orders with positive descriptions of what the shot should do. These were the exact prompt lines:

Playback speed stays exactly constant from the first frame to the last.

The camera keeps travelling through every contact and arrives somewhere new.

It runs beside the two men, slides past a shoulder and re-finds them on the other side.

Both men stay physically engaged for the whole 25 seconds.

Each sentence describes the wanted result directly.

The first fixes playback speed as constant. The second keeps the camera moving through contact instead of lingering or replaying. The third gives the camera a physical route. The fourth keeps both fighters engaged for the full shot.

Those lines also support the pacing decision. If the camera must keep traveling and both men must stay engaged, every mark has to live inside continuous movement. The shot cannot stop to explain what just happened.

With four wounds, I was giving those events less time while asking the camera to pass through all of them. With three marks, each block had enough room for the action to read without abandoning the constant movement rule.

The wording of the marks had already failed moderation

The old version described the wounds with phrases including:

A clean cut opens

fresh blood

the wound reveal

That vocabulary had caused seven out of seven plates in an earlier batch to be rejected by moderation.

The revision changed the setup. It declared at the top that the marks were production makeup. Then it described each mark by pointing to its plate, without naming blood or wounds opening.

This imposed another useful limit. The fight did not need more graphic wording. It needed clear references and readable timing.

Reducing the marks from four to three kept the sequence focused on the events that could be described through their existing plates. It removed pressure to make each contact louder through wording that had already failed.

The fight needed progression, not repetition

The earlier fight repeated instead of rising.

The revision gave every mark a directional consequence: the person who receives it advances rather than retreats.

Toshiro pushes and gains ground, then marks. Sadao turns the fight around and marks. The third arrives when both are already broken, without either man stopping.

That structure gives the three marks different positions inside the fight. They are not separate versions of the same exchange. Each one changes who is carrying the movement.

I described the correction this way:

Ahora, **después de cada marca el que la recibe
AVANZA en vez de retroceder**. Toshiro empuja y gana terreno, marca. Sadao le da la vuelta y
marca. La tercera cae con los dos ya rotos y sin parar.

This was the real use of the four blocks. They did not exist merely to divide 25 seconds evenly. They carried the change in pressure through the fight.

One fighter gains ground. The other reverses it. The last mark lands after both have been worn down, and the movement continues.

The timing and the escalation became the same structure.

What done looked like

The shot was not done when all four original wounds had been squeezed into the prompt. It was done when the remaining three marks had enough time to read, the four blocks carried progression, and both fighters stayed engaged for the full 25 seconds.

It also meant the essential instructions had to remain inside the web interface’s cutoff. Negative orders had to become positive motion. The mark descriptions had to point to plates without repeating the wording that moderation had rejected.

Those constraints are connected. A crowded action plan creates pressure everywhere else. The prompt gets longer. Each event gets less screen time. The camera has less room to move through contact. The marks need increasingly blunt descriptions just to distinguish them.

Removing one wound gave the whole shot room.

Today, divide your shot duration by its action blocks before polishing the prose. If the result gives each event only 4.1 seconds and your own material shows that 6.25 seconds is where the action reads, cut an event before adding more instructions.

---

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