AI video-generation API comparison [2026] — the price of one 8-second 720p video across 8 models

The same 8-second, 720p clip ranges from $0.18 to $3.03 depending on which model generates it—a roughly 16.8-fold difference. This is across 8 video models available on one API key. But there is one thing to know before looking at that number: these 8 models are not billed the same way—3 charge a fixed fee per video, while 5 charge per second of output. So the “cheapest model” changes depending on the length of the clips you actually make. A comparison table listing one number per model without stating the assumed clip length shows a ranking that is only accurate for that duration.

First, our position: Kunavo is our own product, a gateway that resells access to models, not an independent review site. The figures below are read at page render time from the catalog that actually bills requests, so they are real billed rates, not prices copied from announcement materials. Kunavo rates last verified: October 9, 2026.

Price per 8-second, 720p clip, by model

We use an 8-second, 720p clip as a common benchmark because it is the length billed as one unit by per-clip models, and every per-second model supports it. The “Provider list price” column shows the price the provider itself publishes for the same conditions—Veo 3.1 models, as shown in the Google Veo API documentation, and Wan 2.7, as shown in the Alibaba Cloud Model Studio documentation.

Rows with “—” in the provider list price column mean we could not verify a price published by that provider. Therefore, we make no discount claims at all for those models. For the 4 models with a published list price (Google Veo 3.1 Lite・Fast・Quality、Alibaba Wan 2.7), Kunavo's price is lower. For the 4 ByteDance models without a published price, this does not mean they are cheaper than using them directly, and we do not claim that. We prefer leaving cells blank to filling them with estimates.

Two billing methods break comparison tables

Google Veo 3.1 models charge a fixed price per video, and a video is always 8 seconds long. That does not change whatever length the request asks for. ByteDance Seedance models and Alibaba Wan 2.7 charge per second of output. Any comparison table with one number per model silently assumes a duration, and its ranking holds only for that duration.

The consequence is concrete. At 8 seconds, Veo 3.1 Lite tops the table at $0.18. Ask for 4 seconds and all per-second models cost half as much, while Veo models still generate a whole 8-second clip and charge for it—so a per-second model that looked expensive at 8 seconds can become the cheaper choice at 4 seconds. Past 8 seconds, Veo models need another whole clip while per-second models charge only for the extra seconds, so per-second billing gains ground again. At 16 seconds every model costs twice its 8-second price, and the ranking is back to what it was at 8 seconds. Before comparing prices, decide on the clip length your product will actually generate. The AI video cost calculator does that calculation for you.

Which one to choose for each use case

• Drafts, previews, and high-volume production—Veo 3.1 Lite, $0.18 for 8 seconds. At this price, rejecting a clip costs less than the back-and-forth of asking someone whether it is okay to generate it.
• A customer-facing shot—Veo 3.1 Quality (8 seconds: $1.60). The most expensive per-clip option, worth paying for when the output will be used as delivered.
• Short clips around 6 seconds—per-second models. You pay only for what you make. The lowest per-second rate among them is Seedance 2.0 Mini ($0.0526 / second).
• Animating existing artwork—Google Veo 3.1 Lite・Fast・Quality、ByteDance Seedance 2.0 Mini・2.5. It accepts first and last frame specifications and reference images, not just a single image to drive the video.
• Text-only at 1080p—Wan 2.7. It is the only model here without an image-to-video path that also has a published price for comparison.

What Kunavo does not offer

We want to state upfront which of the models commonly grouped together on other pages in these search results Kunavo does not offer. As of September 2026, Sora、Kling、Hailuo、Runway、Luma、Pika is not available. If one of those is on your shortlist, you cannot call it with this key, and another provider is the right choice. Kunavo has a separate guide to using Sora, which says the same thing and explains what alternatives are available.

How to call it

8 models use a single endpoint. Only the model ID changes between models. Generation is asynchronous: the initial request returns a task ID, and the completed video is retrieved in a later poll. Since rendering one clip takes tens of seconds, providers in this space all use the same structure. The request format and polling are covered in the video endpoint documentation, with model-specific details in Veo 3.1 pricing and how to use Seedance. For image generation on the same key, see the image generation API comparison.

Frequently asked questions