Media Transformations Binding for Workers

Cloudflare has introduced a new Media Transformations binding for Workers, allowing developers to perform video operations within their serverless applications. This binding provides a simple API to transform videos stored in any location, including private sources like R2 buckets.

Key Capabilities

The binding supports multiple transformation operations:

• Video Processing: Resize and crop videos with configurable dimensions
• Frame Extraction: Extract still frames for classification or analysis
• Audio Extraction: Pull audio tracks from videos (M4A format)
• Spritesheet Generation: Create multiple frames in a single output

Getting Started

To use the Media Transformations binding, add it to your wrangler.toml or wrangler.jsonc:

[media]
binding = "MEDIA"

Then access it in your Worker code:

const result = env.MEDIA.input(video.body)
  .transform({ width: 480, height: 270 })
  .output({ mode: "video", duration: "5s" });

Use Cases

Developers can leverage this binding for:

• Protected Content Processing: Transform videos stored in private locations without exposing them
• Video Optimization: Resize and compress videos before storing them back to R2 for reuse
• AI Integration: Extract frames for Workers AI vision models or audio for transcription services

Output Formats

The binding supports multiple output modes: video for optimized MP4 clips, frame for single still images, spritesheet for batched frame output, and audio for M4A audio extraction.