How to use Wan API effectively

Wan API usage guide

Wan API is built for generating short video clips from text, images, or multimodal references through either the Playground or the API. The main choice a reader faces is which model fits the material at hand: Wan 3.0 and Wan 3.0 Prime support output up to 1080p and clips up to 30 seconds, while Wan 3.0 Pro and Wan 3.0 Prime Pro add 2K and 4K output options. Wan 3.0 also supports text-to-video, image-to-video, and reference-to-video workflows, and it can generate sound with the video.

The API path starts from a fixed base URL, https://api.wanapi.app. After signing in with Google, adding funds, and creating a key in API key settings, requests use an Authorization: Bearer YOURAPIKEY header. The integration guide covers submitting generation jobs, checking task status, retrieving results, and handling failures, so the documented flow is to authenticate, send a request, poll the task, and download the finished video. Before a first request, the docs also point users to the model reference to confirm the model ID, supported input format, duration limits, output options, and pricing for the model they plan to use.

Input choice matters because the service separates prompt-only, image-started, and reference-driven generation. A text prompt can define a scene, an image can serve as the starting point, and reference media can help guide subject, motion, and mood. The Wan 3.0 guidance uses that same logic in its examples: it suggests describing a single clear moment for text-to-video, using an existing image when you want to animate a still, and assigning roles to image, video, and audio references when the shot needs several kinds of source material. That makes the service useful when you already know what should anchor the clip and want the model to work from that material rather than from a blank prompt alone.

Pricing follows the selected resolution and the combined duration of reference video plus generated output. Text and image inputs do not add video seconds, and turning off generated audio does not lower the per-second rate. The pricing page shows current per-second rates for each Wan model, so the practical cost choice is usually between the output resolution you need and any reference video time included in the request. A 5-second reference video plus a 5-second output is billed as 10 seconds.

For a first test, the documented advice is to keep the scene simple enough to judge motion, subject consistency, and sound. That is the fastest way to see whether the chosen model, input type, and output resolution fit the clip you want before moving to longer or more complex generations.

Back to Wan API