AI video-generation API comparison (2026) — price an 8-second 720p video across 8 models

The price of the same 8-second, 720p video varies by about 16.8 times, from $0.18 to $3.03, depending on the model. That covers 8 video models on the same API key. But before looking at that number, there is something else to know: The 8 models do not all bill the same way—3 charge a fixed amount per video, while 5 charge per second of output. So the “cheapest model” depends on the length of the videos you actually generate. Comparison tables that list only one number per model never say what length they assume, and their ranking is valid only at that length.

Let's be clear about our position. Kunavo is our own product: a gateway that resells access to models, not an independent review site. The figures below are read from the live billing catalog when this page renders, so they are amounts you would actually be charged, not rates copied from an announcement. Kunavo rates last verified: October 1, 2026.

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

We use 8 seconds at 720p as a common baseline because it is the unit length for per-video models and a supported length for every per-second model. The “Provider list price” column shows the provider's own published price for the same conditions: the Veo 3.1 series price is from Google's Veo API documentation, and the Wan 2.7 price is from Alibaba Cloud Model Studio documentation.

Rows with “—” in the list price column mean we could not find a published list price from that provider, so we make no savings claim for those models. For the 4 models with a list price (Google Veo 3.1 Lite／Fast／Quality、Alibaba Wan 2.7), Kunavo's price is below that list price. For the 4 ByteDance models without a list price, that does not mean they are cheaper than going direct, and we make no such claim. We'd rather leave a cell empty than fill it with an estimate.

How two billing methods break comparison tables

The Google Veo 3.1 series charges a fixed amount per video, and every video is fixed at 8 seconds, whatever length the request asks for. The ByteDance Seedance series and Alibaba Wan 2.7 charge per second of output. Any comparison table that lists just one number per model silently picks a video length, and its ranking is valid only at that length.

The result is concrete. At 8 seconds, Veo 3.1 Lite ranks first in the table at $0.18. Ask for just 4 seconds, and the price of every per-second model is cut in half while the Veo series still generates a whole 8-second video and charges for a whole video—so the per-second model that looks expensive at 8 seconds may be cheaper at 4 seconds. Past 8 seconds, the Veo series has to generate another whole video while the per-second models charge only for the extra seconds, so the per-second models gain 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 how long the videos your product will actually generate need to be. You can let the AI video cost calculator do the math.

Which model to use for which situation

• Drafts, previews, and production runs — Veo 3.1 Lite, $0.18 for 8 seconds. At this price, the cost of discarding a video is less than the back-and-forth of asking someone, “Can this be generated?”
• The one shot you're delivering to a client — Veo 3.1 Quality ($1.60 for 8 seconds). The most expensive per-video option, and worth paying for when the result goes straight to the client.
• Short videos around six seconds — a per-second model, where you pay for what you generate. The lowest price per second is Seedance 2.0 Mini ($0.0526/sec).
• Animating existing artwork — Google Veo 3.1 Lite／Fast／Quality、ByteDance Seedance 2.0 Mini／2.5, which accept start and end frames and reference images, rather than just a single driving image.
• Text only, with a 1080p requirement — Wan 2.7. The only model here with a published list price to compare against that has no image-to-video path.

What Kunavo does not offer

Let's start with other models listed together on pages in these search results that Kunavo does not offer. As of September 2026, Kunavo does not offer Sora、Kling、Hailuo、Runway、Luma、Pika. If one of these is on your shortlist, this API key cannot call it; the answer is another provider.

How to call the API

All 8 models share one endpoint; only the model ID changes between them. Generation is asynchronous: the first request returns a job ID, and the completed video is retrieved in a later poll. A video takes tens of seconds to generate, so providers in this field all use this structure. Request formats and polling are covered in the video endpoint documentation, with model-specific details in the Veo 3.1, Seedance, and Wan 2.7 guides. For image generation on the same API key, you can try a cost estimate with the AI image generation cost calculator.

Frequently asked questions