AI video API for every top model

One AI video API, 9 models

Every model below takes the same request shape — model, prompt, duration, resolution, aspect ratio and audio — and returns the same video object. Switching from one model to another is a one-word change in your code.

“From” is the shortest clip at the lowest resolution. Longer clips, higher resolutions and sound cost more; the request builder above shows the exact figure.

How the SI video API works

1. Create a key. Sign in to the dashboard with Google or an email link and create a key. It is shown once; store it as a secret on your server.
2. Create a video. POST /v1/videos with a model and a prompt. The response arrives immediately with an id, the status pending and the credit cost.
3. Poll and download. GET /v1/videos/{id} until the status is succeeded, then fetch video_url. For image-to-video, upload the frame to /v1/uploads first and pass its id.

Why one AI video API instead of each vendor

Calling every video model directly means a separate account, contract, billing setup and request format for Google, ByteDance, Kuaishou, Alibaba and the rest — and rewriting your integration whenever a better model appears. With SiVideoAPI you integrate once. New models show up in /v1/models and work with the code you already have.

• One balance: buy credits once and spend them on any model, without minimums per vendor.
• Normalized parameters: durations, resolutions and aspect ratios use the same names for every model.
• Refunds built in: if a provider errors out or blocks the output, the credits return without a support ticket.
• Stable URLs: finished clips are copied to our own storage, so your links never depend on a provider’s temporary files.

What teams build with an AI video API

• Ad and product variations: render the same product shot in ten settings and test which one sells, without a film crew.
• Social content pipelines: turn a daily script or a blog post into short vertical clips on a schedule.
• Apps with a “make a video” button: let your own users animate a photo or an idea, and pass the cost through to them.
• Game and story prototypes: generate cutscenes, trailers and mood reels before committing to production.
• Internal tools: explainer clips for training material, real-estate walkthrough drafts, event teasers.

In each case the AI video API does the rendering while your product keeps the workflow, the branding and the customer relationship. Because every call returns its exact cost, it is easy to put a price on a feature before you ship it.

Built for production use

The API runs on Cloudflare’s global network in front of fal.ai’s GPU fleet. Each account can run up to ten renders at once and send twenty requests a minute; ask us if you need more. Prompts pass an automated safety check before any credits are taken, so a blocked prompt costs nothing. Videos stay available until you delete them with DELETE /v1/videos/{id}, and reference images you upload are removed after thirty days. Webhooks are not available yet, so plan to poll.

Pricing in brief

The AI video API is billed per second of output in credits, the same currency for every model. Packs start at $9.90 for 1,100 credits, larger packs add bonus credits, and nothing renews. Every video object includes cost_credits, and GET /v1/balance returns what is left. The full table is on the video API pricing page.

SI or AI?

Since September 2026 the US government calls this technology SI, short for super intelligence. We use the new name, but the models are the ones you know: if you were looking for an AI video API or a Veo 3 API, you are in the right place. See the Veo 3 API page for Google’s model in detail.

Questions about the AI video API

What is an AI video API?

It is a web API that turns a text prompt, or a prompt plus an image, into a video clip. You send an HTTP request, the model renders the clip on GPU servers, and you download an MP4. SiVideoAPI puts several leading video models behind one such API.

Which video models can I call?

9 models from Google, Kuaishou, MiniMax, Alibaba, ByteDance, xAI. GET /v1/models returns the current list with durations, resolutions, aspect ratios, audio support and the credit price per second.

How much does the AI video API cost?

You pay per second of video in credits; the cheapest clip today is 72 credits on Veo 3.1 Lite. Credit packs start at $9.90 for 1,100 credits, there is no subscription or monthly minimum, and credits never expire.

Is there a free trial?

New accounts that sign in with Google get 80 free credits, enough for a short test clip on the cheapest settings. No card is needed to create a key.

Do you support webhooks?

Not yet. Create a video, then poll GET /v1/videos/{id} every five to ten seconds until the status is succeeded or failed. Most clips finish in one to three minutes.

What happens when a render fails?

The video is marked failed with a readable error, and its credits go back to your balance automatically. That includes provider errors, safety-filter blocks and renders that time out.

Need higher limits or a model we don’t list? Email support@sivideoapi.com.