How long a generation takes, and who puts a number on it
Four channels here say something about how long work takes. Two label plans slower, faster and fastest, one reserves a priority queue for a higher plan, and one publishes measured median times for each step of a production and states that paying more does not make a single generation faster. As of 2026-10-13.
| Channel | What is published about time | In what form |
|---|---|---|
| Hedra | Slower generations on Basic, faster on Pro, fastest on Ultra | Comparatives by plan |
| HeyGen | Standard, fast, faster and fastest video processing by plan | Comparatives by plan |
| Pika | A priority queue on one higher plan only | A queue benefit |
| SceneMixer | Medians: about 3 min 20 s a segment, under 4 min to composite an episode | Measured figures, per step |
Inclusion rule. Channels whose price or product pages address how long generation takes or how jobs are queued. A channel that is silent on time is not given a row. Order. Alphabetical by channel.
1Speed sold as a word
Two plan grids sell time without a unit: slower, faster, fastest. A buyer learns the order of the tiers and nothing about the size of the difference, which is the number a schedule needs. A priority queue on one higher plan is the same promise from the other end: someone else waits first.
2Measured figures, step by step
One channel publishes medians taken from every job completed in a thirty-day window: about two minutes forty-nine seconds to analyse a whole script, one minute fifty-four for an episode's storyboard, three minutes twenty for a video segment and three minutes fifty-two to composite an episode. Because segments run in parallel, the same page puts an episode at about a quarter of an hour of machine time.
A median is not a promise, and the page does not present it as one. It is still the only figure here a production can plan a day around.
3What paying more actually changes
The same channel's plan guide is explicit: a plan changes how many jobs run at once, from three on a free account to twenty on the top tier, and does not make any single generation faster, because the same model does the same work on every tier. Its priority queue only decides who waits when the site is busy.
That separates the two things the comparative labels blur. Throughput can be bought; latency, on a given model, mostly cannot.
- Generation speed by planPlans are labelled slower generations on Basic, faster generation on Pro and fastest generation on Ultra
- Generation speed by planPlans list standard video processing on Free, fast on Creator, faster on Pro and fastest higher up
- Queue priority by planA priority queue is marked not included on the lower plans and included on one higher plan
- How long each step takesMeasured over thirty days to 20 September 2026, median times are about 2 min 49 s to analyse a script, 1 min 54 s per storyboard, 3 min 20 s per video segment and 3 min 52 s to composite an episode
- What a plan does to speedA plan does not make a single generation faster, because the same model does the same work on every tier; a priority queue only decides who waits when the site is busy
- ConcurrencyA free account runs three generations at once; the paid tiers run five, eight, ten, fifteen and twenty, and the top three add a priority queuea figure for every tier
4Sources
Each figure is read from the vendor page linked in the ledger, checked 2026-10-13. How a figure qualifies is on method; what a rate leaves out is on limits and unspent credits and parallel generations. Nearby: The cheapest second, Credits vs dollars, What it excludes.