Rendering with Vercel Sandbox
Vercel Sandbox allows you to render videos on-demand without managing Lambda or AWS infrastructure.
Each render spawns an ephemeral Linux VM with full access to the Remotion renderer.
As a reference implementation, we have created a template: Next.js Sandbox Template.
Getting Started
Use the Vercel template:
bash
Push it to Vercel.
Then, in your Vercel dashboard, under "Storage", create a new Blob store and connect it to the project.
Redeploy for the changes to take effect.
Advantages
This is the easiest way to setup Remotion server-side rendering. You only need one Vercel account and connect a blob storage - then you just push to deploy.
Vercel has higher limits than Lambda by default, so you can render more videos at once than with a fresh AWS account.
Allowing for 2000 concurrent sandboxes on the Pro Plan allows for massive scale.
Limitations
Rendering happens on a single machine rather than a distributed system, so it is slower than Lambda.
Spawning a new VM for each render has a setup phase which includes installing dependencies and downloading a browser.
We plan to soon improve this by using Vercel Sandbox snapshots.
Currently, the browser downloaded does not have support for decoding H.264, meaning that using @remotion/media will often fallback to <OffthreadVideo>.
We also plan to resolve this by using a custom build of Chrome soon.
Limits
The timeout limit for Vercel Sandbox is 45 minutes on the Hobby plan, and 5 hours on the Pro and Enterprise plans.
For concurrency, the limit is 10 simultaneous renders on Hobby, and 2000 on Pro and Enterprise.
For more details, see: Vercel Sandbox pricing
Source Code
See the template source code for implementation details.