How many generations run at once on HeyGen
HeyGen describes one tier as having twice the concurrency of the tier below it and gives no figure for either. A ratio without a base is the least usable shape this column takes and it is still an answer of a kind. As of 2026-09-22.
| What is published | As the pricing page states it |
|---|---|
| Concurrency on the higher tier | Described as twice that of the tier below |
| Concurrency on either tier | No figure given for either |
| Longest single video | Up to 30 mins below one tier, up to 60 mins on it, unlimited above |
| Credits a video consumes | Stated to vary by model, duration and complexity |
Inclusion rule. What the pricing page states about parallel capacity, with the figures it does publish for comparison. A multiple counts as a statement; it simply cannot be sized. Order. The ratio first, then the missing base, then what is quantified.
1Where HeyGen sits in this column
| Channel | Parallel generations |
|---|---|
| HeyGen | Twice as much on one tier, no figures |
| Kling AI | Twenty per package, unlimited on membership |
| SceneMixer | 3 free, then 5, 8, 10, 15 and 20 by tier |
| Vidu | Five free, then $2,080 a month each |
Inclusion rule. Channels whose public pages address how many generations may run at the same time. Silence on the subject is the norm and is recorded on the ledger rather than here. Order. Alphabetical by channel name.
2A multiple with no base is upgrade copy doing double duty
Twice as much of something unspecified is a reason to move up a tier and not a specification. It works on a comparison table where every row is a tick or a phrase, and it fails the moment somebody has to decide whether the tier is sufficient.
The shape is common in tier grids and rare in this register, because most pages here either state a number or say nothing. Saying twice is the middle option and it inherits the weaknesses of both.
3The page quantifies length and declines to quantify throughput
Video length is published precisely and per tier: half an hour, an hour, unlimited. So this is not a page that avoids numbers. It has chosen which capacity to express as a figure and which to express as a comparative.
That choice is legible. Length is a product boundary a buyer hits on their first long video; concurrency is an infrastructure fact that moves with capacity planning, and sellers are reluctant to commit to it in writing.
4Why this column is thin across the whole register
Four channels here say anything at all about parallel generation, and three attach a figure. Concurrency is where a seller's spare capacity lives, and publishing it converts a flexible operational choice into something buyers will hold them to.
The result is that throughput stays outside almost every comparison, including the cheap ones. A rate reachable at one job at a time and the same rate reachable at twenty are not the same offer, and nothing in a price table shows the difference.
- Parallel generationBusiness is described as having 2x more concurrency than Pro, with no figure given for eitherpublished as a multiple
- Longest single videoVideos up to 30 mins on the plans below Business, up to 60 mins on Business, with no duration maximum on Enterprisestated per plan
- Why no credit count per video is printedCredits are described as the single currency for video and asset creation, and the number a video uses can vary by model, duration and complexity
5Sources
Read from the pricing page published at heygen.com on 2026-09-22. The rate this column sits beside is on HeyGen; what the register accepts as published is on the method page. The same column on other channels: Vidu, Kling AI. Every column: what each channel publishes.