How long do 100 AI video clips take? Concurrency slots by plan

Sume limits how many paid generation jobs run at once per workspace: 1 on Free, 4 on Pro, 8 on Startup and 20 on Scale. A batch of N equal jobs finishes in ceil(N / slots) waves, so 100 clips take 100, 25, 13 and 5 waves. Total time is waves times the time of one job, T.

Waves by plan

The last column assumes T = 3 minutes purely as an illustration. It is not a benchmark; measure your own T on a handful of jobs and substitute.

Why Startup is 13, not 12

100 / 8 = 12.5, and a partial wave still takes a full job time, so it rounds up to 13. In practice workers claim queued jobs as slots free up, so the real figure is a little better than a strict wave model, but the ceiling is a good planning number.

Slots are separate from price

A slot is not a discount. Each clip still costs the same per second on every plan. What the higher plans buy is parallelism and a deeper queue. Prepaid top-ups do not raise the processing limit; the docs state concurrency is plan-only, with admin overrides for enterprise contracts. Always read generation_limits.concurrency_limit from a submit response rather than hard-coding this table.

import math

SLOTS = {"free": 1, "pro": 4, "startup": 8, "scale": 20}


def main() -> None:
    jobs = 100
    minutes_per_job = 3  # replace with your own measurement
    for plan, slots in SLOTS.items():
        waves = math.ceil(jobs / slots)
        print(plan, waves, "waves", waves * minutes_per_job, "min")


main()

What to do with the number

If the wall clock is the constraint, compare the cost of a bigger plan with the cost of the delay. Going from Pro to Scale takes a 100-clip batch from 25 waves to 5, a 5x speed-up for 10x the monthly fee ($400 against $40), but it also raises included usage from $40 to $800.

The same math for 500 and 1,000

The formula scales. For 500 jobs: Free 500 waves, Pro 125, Startup 63, Scale 25. For 1,000 jobs: Free 1,000, Pro 250, Startup 125, Scale 50. If a job takes 3 minutes, 1,000 jobs on Scale is 150 minutes, 2.5 hours; on Pro it is 750 minutes, 12.5 hours.

These are ceilings from a strict model. Jobs vary in length, and a slot that frees early starts the next job at once, so real batches tend to finish somewhat sooner. The takeaway is the ratio: moving from 4 to 20 slots cuts waiting by a factor of 5, and from 8 to 20 by a factor of 2.5.