If you’re a developer living in 2026, you know the struggle:
You have five different API keys for text, three for images, and now you have to juggle Kling, Veo, and Wan for video.
Vercel has integrated them all into a single, unified Agentic Core
Yes, the Vercel AI Gateway now supports Video Generation.
One SDK to Rule the Pixels
We’ve been living in a fragmented world.
I remember back when I was building my webapp, I had to write custom wrappers for every new model that hit the market.
Vercel AI Gateway has solved this by supporting video generation programmatically through the AI SDK 6.
You use the same interface for a text prompt that you use to generate a cinematic, photorealistic video clip.
No new authentication flows.
No separate observability dashboards.
Just one generateVideo call.
Automating stuff
I tested the Image-to-Video capability tonight on my MacBook. I took a still photo of a football game from one of my database and piped it through klingai/kling-v2.6-i2v.
Within 30 seconds, I had a 5-second cinematic clip of the crowd cheering with realistic motion blur.
I didn’t have to worry about Quota Anxiety because the AI Gateway playground lets you compare providers and tweak prompts without writing a single line of boilerplate.
An example boiler plate code looks like this:
import { experimental_generateVideo as generateVideo } from 'ai';
async function generateWeeklySummary(userPhotoUrl, matchStats) {
console.log("Starting video generation...");
const { videos } = await generateVideo({
// Using Veo 3.1 for high visual fidelity and physics realism
model: 'google/veo-3.1-generate-001',
prompt: {
image: userPhotoUrl,
text: `Cinematic sports highlight. The crowd in the background is cheering.
Slow motion effect on the player. Professional stadium lighting.
Match Context: ${matchStats.result}. High production quality.`,
},
aspectRatio: '16:9', // Perfect for YouTube or LinkedIn shares
providerOptions: {
google: {
// Ensuring we get that high-end cinematic feel
mode: 'pro',
fps: 30
}
}
});
// Exporting to our 'Recaps' bucket
return videos[0].url;
}The Big Four: Kling, Veo, Wan, and Grok
Vercel isn’t just giving us one model, they’re giving us everything.
- Google Veo: This is the physics king. If you need cinematic lighting and realistic physics (like an owl turning its head with intricate feather detail), this is your primary.
- Kling (Kuaishou): The best for “First and Last Frame” generation. You upload an empty room and a furnished one, and it generates the seamless transition in between.
- Alibaba Wan: The specialist for “Reference-to-Video.” It can extract a character from an image and keep that identity consistent across different scenes — perfect for brand mascots.
- xAI Grok Imagine: The “Sprinter.” It’s incredibly fast at style transfer, like turning a source video into a watercolor painting in seconds.
The Prompt Caching Advantage
Just like with the Antigravity update we discussed last week, the real win here is Observability.
When you’re generating video, the costs are high.
Vercel’s Gateway lets you see exactly which models are burning your credits and which ones are delivering the best “Instruction Following”.
If you are a developer in India using a Jio 5G plan, the latency on these new video endpoints is finally becoming manageable.
You can now generate Product Reveals on-demand without paying massive licensing fees.
Now, stop acting as the manual bridge between your images and your videos. Fire up the AI Gateway, claim your compute, and start shipping things that move :)
In case we are meeting for the first time, come over here, it’ll be worth the roller coaster of articles that are gonna come up in the next few weeks.