Cloudflare released Streamline's source code and a developer playground for custom video-processing workflows. Examples add overlays, subtitles and filters to streams. A media engine runs inside a container, with Workers controlling sessions and Durable Objects coordinating them. Processing can continue through a control connection's interruption; the current system uses CPUs, with higher-resolution work facing capacity constraints.
Cloudflare released Streamline, a developer playground and open-source example for custom video-processing workflows. It demonstrates adding overlays, burned-in subtitles and filters to streams. The project brings together Cloudflare's Stream video service, its Workers application platform, Containers for running software in isolated environments and Durable Objects for coordinating sessions.[1], [2]
The media engine runs inside a container. A Go control layer translates incoming requests into operations, and the current processor uses FFmpeg, a tool for handling video and audio. A Worker starts, configures, monitors and stops sessions. A Durable Object coordinates the session and the preview relay.[1]
Processing can continue when the controlling application disconnects, and the client can reconnect to resume control. A maximum duration prevents an abandoned session from running indefinitely. Local development can use Docker without a Durable Object in the single-user arrangement.[1]
Cloudflare published the engine and demonstration application in separate GitHub repositories. The public playground permits one active session per user and shares capacity across sessions. Its current CPU-based processing can become a bottleneck at higher resolutions and frame rates. Private deployments use Cloudflare Access for authentication and bind session control to a verified user.[1]
