This example demonstrates how to use Workflow DevKit with Express and Vercel Sandbox to build an audio compression service. Audio files are uploaded via multipart/form-data, processed by FFmpeg within a durable workflow running in an isolated sandbox, and streamed directly back in the HTTP response.

• Sandbox isolation: FFmpeg runs in a Vercel Sandbox for secure, isolated processing
• Workflow orchestration: Audio compression is broken into multiple steps (createSandbox -> setupFfmpeg -> transcode -> streamOutput -> stopSandbox)
• Streaming support: Results are streamed back to the client as they're produced
• FFmpeg compression: Converts audio to AAC codec in M4A container at 128kbps

• VERCEL_OIDC_TOKEN: The workflow runtime expects the VERCEL_OIDC_TOKEN environment variable to be present for @vercel/sandbox. When running inside a Vercel Sandbox this is injected automatically; if you run the server outside of Vercel, you must provide a valid OIDC token yourself.

1. Clone this example and install dependencies:

   git clone https://github.com/vercel/workflow-examples
   cd workflow-examples/ffmpeg-processing
   pnpm install

2. Link your Vercel project:

   npx vercel link

3. Fetch the VERCEL_OIDC_TOKEN:

   npx vercel env pull

4. Start the development server:

   pnpm dev

5. Test the audio compression workflow:

   # Convert a WAV file to compressed M4A
   curl -X POST -F "file=@input.wav;type=audio/wav" -H "Expect:" http://localhost:3000/convert --output output.m4a

   The endpoint accepts any audio file and returns a compressed M4A file. Be sure to update the audio file extension if not a .wav file.

Compresses an uploaded audio file using FFmpeg within a workflow.

Request:

• Content-Type: multipart/form-data
• Field name: file
• Accepted formats: WAV, MP3, OGG, FLAC, AAC, M4A

Response:

• Content-Type: audio/mp4
• Body: Compressed audio file (M4A/AAC at 128kbps)

Example:

curl -X POST -F "file=@podcast.wav;type=audio/wav" -H "Expect:" http://localhost:3000/convert --output podcast.m4a

1. Express route receives the upload via multer.memoryStorage()
2. Workflow orchestrates five steps:
   • createSandbox: Provisions a Vercel Sandbox instance
   • setupFfmpeg: Downloads and installs FFmpeg in the sandbox
   • transcode: Writes input to sandbox, runs FFmpeg to compress audio
   • streamOutput: Streams the compressed file back to the workflow
   • stopSandbox: Cleans up the sandbox instance
3. Response streams the compressed bytes directly to the client

ffmpeg-processing/
├── src/
│   ├── index.ts                          # Express app with /convert route
│   └── workflows/
│       └── audio-convert/
│           ├── index.ts                  # Main workflow orchestration
│           └── steps/                    # Individual workflow steps
│               ├── create-sandbox.ts
│               ├── setup-ffmpeg.ts
│               ├── transcode.ts
│               ├── stream-output.ts
│               └── stop-sandbox.ts
├── nitro.config.ts                       # Nitro configuration with workflow module
├── package.json
├── tsconfig.json
└── README.md

• Workflow DevKit Documentation
• Express Getting Started Guide
• FFmpeg Documentation