On this page
The ClipCost AI video price index is a count of every active, sourced per-second price for AI video generators, rebuilt from one open dataset. The headline: how much AI video costs per second depends on the model far more than on the average.
From the ClipCost data (154 active per-second prices for 37 models on 17 routes, checked Nov 5, 2026): median $0.13 per second, so an 8-second clip at the median costs $1.04; the middle half runs $0.084 to $0.231 per second; cheapest Seedance 1.0 Pro Fast on BytePlus ModelArk ($0.01), most expensive Seedance 2.5 (US endpoint) on fal.ai ($1.16).
Last checked . Prices are re-checked every week (planned: Wednesdays). Each row shows its own check date; a row older than 14 days is out of date.
This page is the summary and the citable numbers. The full sortable list lives on the live price table, and dated changes are on price history.
Key findings
The market in one line
The line above is the whole index: the median sets the typical price, and the spread between cheapest and most expensive shows how much the choice of model matters. The cheapest and the dearest entries differ by two orders of magnitude, so an average hides more than it tells.
By model family
| Family | Prices | Median per second | Range | Cheapest |
|---|---|---|---|---|
| Hailuo | 12 | $0.0615 | $0.025 - $0.16 | MiniMax H3 Lite, MiniMax API |
| Runway Gen | 5 | $0.0778 | $0.05 - $0.12 | Gen-4 Turbo, Runway API |
| Grok Imagine | 11 | $0.08 | $0.02 - $0.25 | Grok Imagine Video 1.5 Lite, xAI API |
| Wan | 11 | $0.10 | $0.05 - $0.28 | Wan 3.0 (wan3.0-video), Alibaba Cloud Model Studio |
| Gemini Omni (Veo successor) | 1 | $0.101 | $0.101 - $0.101 | Gemini Omni 1.1 Flash, Gemini API |
| Kling | 37 | $0.126 | $0.0615 - $0.42 | Kling 4.0 Flash, Kling API |
| HappyHorse | 5 | $0.14 | $0.07 - $0.18 | HappyHorse 1.1, Alibaba Cloud Model Studio |
| LTX | 16 | $0.145 | $0.03 - $0.39 | LTX-2.3 Fast, LTX API |
| Seedance | 32 | $0.154 | $0.01 - $1.16 | Seedance 1.0 Pro Fast, BytePlus ModelArk |
| Veo | 23 | $0.28 | $0.08 - $0.60 | Veo 3.1 Fast, Vertex AI |
| Runway Aleph | 1 | $0.28 | $0.28 - $0.28 | Aleph 2.0, Runway API |
Active per-second prices at every listed resolution, with and without audio, checked Nov 5, 2026.
Hailuo, Runway Gen-4 and Grok Imagine sit at the low end of the medians, while Veo and Runway Aleph sit at the top. Kling and Seedance have the most listed prices because they sell several tiers, resolutions and routes. A wide range inside one family usually means a resolution or audio step, not a different product.
Cost of one 8-second clip
The chart takes the base variant of each family on its cheapest route. Compare families only at the same resolution and audio setting, which the cost-per-clip guide walks through.
720p and API only
From the ClipCost data (58 active per-second prices for 28 models on 11 routes, checked Nov 5, 2026): median $0.12 per second, so an 8-second clip at the median costs $0.96; the middle half runs $0.0923 to $0.155 per second; cheapest Seedance 1.0 Pro Fast on BytePlus ModelArk ($0.0206), most expensive Seedance 2.5 (US endpoint) on fal.ai ($0.473).
From the ClipCost data (97 active per-second prices for 32 models on 9 routes, checked Nov 5, 2026): median $0.12 per second, so an 8-second clip at the median costs $0.96; the middle half runs $0.07 to $0.18 per second; cheapest Seedance 1.0 Pro Fast on BytePlus ModelArk ($0.01), most expensive Seedance 2.0 on BytePlus ModelArk ($0.78).
Restricting to 720p narrows the spread compared with the full index, because the extremes of the full index are removed. Pay-as-you-go APIs are the largest group and their median is close to the all-routes figure.
Resellers against direct
From the ClipCost data (17 model and resolution pairs sold both direct and through fal.ai or Replicate, checked Nov 5, 2026): the median reseller price equals the direct API price, from -33% to +50%. 9 pairs cost the same, 4 are cheaper through the reseller and 4 cost more.
Reseller listings are not a uniform premium: some models cost the same through fal.ai or Replicate, some cost more, a few cost less. See fal.ai vs Replicate vs direct API for the model-by-model comparison.
Method
- What is counted: active per-second prices only, at every listed resolution, with and without audio. Per-clip prices and credit-based app plans are not mixed into these figures. Plans are listed separately in the plans file and appear, flagged as derived, on the live table.
- Which price: the list price in US dollars as printed by the provider or reseller, before tax and before any retries. Temporary promos are not applied.
- Derived values: where a provider prints units or tokens instead of dollars, the dollar price is computed and flagged as derived. Formulas are on the methodology page.
- Excluded: shut down and legacy models, and items without a verified price. Unknown values stay empty and are never estimated.
- Check date and cadence: the date in each figure above is the date of the last data check. Prices are re-checked weekly and every row keeps its own check date.
Prices change often. Confirm on the provider's page before you pay.
What changed since the last check
This is the first edition of the index, dated 5 Nov 2026, so there is no earlier edition to compare with. Price moves are logged from now on on price history, and this page is re-run with each weekly check.
Use the data
Download the data: ai-video-prices.csv ยท offers.json
Free to reuse under CC BY 4.0: credit ClipCost with a link to this page. Each row keeps its source and check date.
Cite the index as "ClipCost AI video price index" with a link to this page. For your own numbers, use the cost calculator to turn a per-second price into a cost for a clip length and retry rate you choose.
Questions people ask
How much does AI video cost per second?
The index headline above gives the median across all active per-second prices, together with the interquartile range and the cheapest and most expensive entries. The family table shows the median for each model family.
Which AI video model is cheapest per second?
The cheapest entry in the index is named in the headline, and each family row names its own cheapest model and route. Cheap entries are often 480p or silent, so compare at the same resolution and audio setting.
Do resellers like fal.ai and Replicate cost more than the maker?
It depends on the model. The reseller line in the findings reports how many model and resolution pairs cost the same, less or more compared with the direct API.
Sources
Prices and limits were read from the sources below on the dates shown. They change often: confirm on the provider's page before you pay.
- Google Cloud: Veo pricing - 2026-11-05
- Kling API pricing - 2026-11-05
- BytePlus ModelArk pricing - 2026-11-05
- Runway API pricing - 2026-11-05
- MiniMax pay-as-you-go pricing - 2026-11-05
- xAI API pricing - 2026-11-05
- ClipCost open dataset (JSON) - 2026-11-05