learnGetting started
How many words fit in one AI video shot: a script budget per engine, with credit cost
AI video engines cap every shot at a fixed length, so your script has a word budget per shot. The words that fit in 4, 6, 8 and 15 seconds, and what each line costs.
by reeldouble · October 28, 2026 · 7 min read
The 150-words-a-minute rule is right, and not enough
Every guide to script length says the same thing: people speak about 150 words a minute, roughly 2.5 words a second, so a 60-second video needs about 150 words. Slower, more emotional delivery runs closer to 2.2 words a second. All true.
What those guides are written for is a talking head reading one long script. An AI video is not that. It is a string of shots, and every engine caps how long one shot can be. Veo 3.1 Fast renders 4, 6 or 8 seconds and nothing in between. Kling O3 Pro renders up to 15 seconds in one shot. A line of dialogue that runs 9 seconds does not fit in a Veo shot at all, and a line that runs 3 seconds is still billed as 4, because 4 seconds is the shortest render anybody pays for.
So the useful unit is not words per minute. It is words per shot, and every shot has a price. This is how we budget a script before anything renders.
The word budget per shot length
Using 2.5 words a second for brisk delivery and 2.2 for slower, and leaving about half a second of air at each end so the line does not start on the first frame:
| shot length | spoken time | words, brisk | words, slow | good for |
|---|---|---|---|---|
| 4 seconds | about 3 seconds | 7 to 8 | 6 to 7 | a punchline, a reaction, a single question |
| 6 seconds | about 5 seconds | 12 to 13 | 11 | one full sentence of dialogue |
| 8 seconds | about 7 seconds | 17 to 18 | 15 | a sentence and a short follow-up |
| 12 seconds | about 11 seconds | 27 | 24 | a short exchange or a narrated beat |
| 15 seconds | about 14 seconds | 35 | 30 | a narrated passage, the longest single Kling shot |
The half-second of air matters more than it looks. Engines need a beat for the character to react before speaking and after finishing. Lines that fill the shot edge to edge tend to come back rushed or clipped.
Which lengths each engine serves
The word budget only helps if the engine actually renders that length. These are the limits our studio enforces:
| engine | lengths it renders | one shot holds up to |
|---|---|---|
| Veo 3.1 Fast | 4, 6 or 8 seconds only; always 8 with a character reference | about 17 words |
| Kling O3 Pro | 4 to 30 seconds, as shots of up to 15 | about 35 words per shot |
| Seedance 2.0 and 2.0 Fast | 4 to 15 seconds | about 35 words |
| Seedance 2.5 | 4 to 30 seconds in one take | about 70 words |
Veo's rule is the one that shapes scripts the most. If you are using a reference image to hold a character on Veo 3.1 Fast, every render is 8 seconds, so every line you write for it should be about 15 to 18 words. Shorter lines leave dead air you are paying for. Longer lines do not fit.
What each line costs
Once a line has a length, it has a price. These are the credits our studio charges for one voiced shot at 720p:
| shot | Kling O3 Pro | Veo 3.1 Fast | Seedance 2.0 | Seedance 2.5 |
|---|---|---|---|---|
| 4s, about 8 words | 120 | 120, or 240 with a reference | 180 | 280 |
| 6s, about 13 words | 180 | 180, or 240 with a reference | 270 | 420 |
| 8s, about 18 words | 240 | 240 | 360 | 560 |
| 15s, about 35 words | 450 | not served | 675 | 1,050 |
Two things to notice. On Kling O3 Pro and Veo 3.1 Fast a voiced second is 30 credits, so every word you cut from a long line saves about 12 credits. And shots with no dialogue can run silent: 24 credits a second on Kling O3 Pro, 20 on Veo 3.1 Fast. Seedance bills the same whether there is sound or not, so silence is not a saving there.
At the Lite rate of 1,000 credits for $10, a 13-word line on Kling O3 Pro is $1.80.
A 60-second script as a shot budget
Here is how we would lay out a 60-second story with one character, about 140 words of dialogue and narration, on Kling O3 Pro:
| shot | length | words | voiced or silent | credits |
|---|---|---|---|---|
| 1. hook line | 4s | 8 | voiced | 120 |
| 2. setup | 8s | 18 | voiced | 240 |
| 3. establishing, no dialogue | 4s | 0 | silent | 96 |
| 4. problem | 8s | 17 | voiced | 240 |
| 5. reaction | 4s | 6 | voiced | 120 |
| 6. turn | 12s | 26 | voiced | 360 |
| 7. payoff, no dialogue | 4s | 0 | silent | 96 |
| 8. resolution | 8s | 18 | voiced | 240 |
| 9. closing line | 8s | 17 | voiced | 240 |
| total | 60s | 110 spoken | 1,752 |
That is 110 words spoken across 60 seconds, well under the 150 the generic rule allows, and that is deliberate. AI shots read better with room in them, and two silent shots save 24 credits each against voicing them. The same 60 seconds all voiced would be 1,800 credits on Kling O3 Pro. On Seedance 2.5 at 720p it would be 4,200.
How we cut a script to fit
- Write the script as prose first. Do not think about shots yet.
- Read it aloud with a timer. Real timing beats any words-per-second rule, especially for jokes and pauses.
- Break it at every change of speaker or idea. Each piece is a candidate shot.
- Check each piece against the table above. Anything over about 35 words gets split. Anything under about 7 words gets merged with its neighbour or given a silent shot.
- Mark the silent shots. Establishing frames, reactions, transitions. On Kling and Veo they are cheaper.
- Total the credits before rendering. If the total is over budget, cut words from the longest lines first, because those are the most expensive seconds.
Our studio's writer does step 3 for you when you paste a long script, breaking it into a scene list, and every render quotes its credits before you press the button. A writer run is 5 credits; turning an overlong script into scenes costs nothing. You can paste a script into the studio and see the per-shot quote straight away.
Three mistakes we see in AI video scripts
Lines that are too long for the engine. A 25-word line on Veo 3.1 Fast will not fit into 8 seconds at a natural pace. The engine either rushes it or cuts it. Split it across two shots.
Every shot voiced. A story where someone talks the entire time is tiring to watch and costs more. One or two silent beats per minute give the viewer room and save credits on Kling and Veo.
Writing for the edit, not the render. If you plan to trim a 6-second shot to 3 seconds in the edit, you are still paying for 6. Write the line to the length you will use, and pick the shortest length the engine serves that fits it.
For the bigger picture of turning a script into a finished video, the script to video guide walks through the whole pipeline, and the 60-second shot split guide covers where to cut by story beat rather than by word count.
FAQ
How many words can I fit in an 8 second AI video clip?
About 15 to 18 words at a natural pace, leaving half a second of air at each end. That is the most useful number for Veo 3.1 Fast, which renders 4, 6 or 8 seconds and always 8 when a character reference is attached on ReelDouble.
How many words is a 60 second AI video script?
The usual rule is 150 words, but AI video reads better at 110 to 130 spoken words with a couple of silent shots. On ReelDouble a 60-second Kling O3 Pro story with two silent shots costs about 1,750 credits.
How long can one AI video shot be?
It depends on the engine. On ReelDouble Veo 3.1 Fast renders 4, 6 or 8 seconds, Seedance 2.0 up to 15, Kling O3 Pro up to 15 per shot and 30 per render, and Seedance 2.5 up to 30 seconds in one take.
How much does one line of AI dialogue cost?
On ReelDouble a 13-word line in a 6-second voiced shot is 180 credits on Kling O3 Pro or Veo 3.1 Fast, 270 on Seedance 2.0 and 420 on Seedance 2.5 at 720p. At the Lite rate that Kling line is $1.80.
Should every AI video shot have dialogue?
No. One or two silent shots per minute make a story easier to watch, and on ReelDouble they are cheaper on Kling O3 Pro, 24 credits a second instead of 30, and on Veo 3.1 Fast, 20 instead of 30.
