SNACK: 3-line summary
- Cloudflare released Streamline, an open-source developer playground for custom video processing.
- Its examples cover webcam picture-in-picture, image overlays, filters and burning existing subtitles into video.
- Developers can run it locally or in their own account; the hosted demo has access and session limits, and CPU processing can bottleneck demanding output.
Cloudflare released Streamline on October 2, 2026 as an open-source developer playground for custom video processing. Its examples show how to add graphics to webcam or live feeds and burn existing subtitles into hosted videos.
Snackgirls react
Red: I’d layer my webcam over a game livestream and swap the overlay when a match gets tense. That’s the kind of setup I’d like to tinker with.
AIKO: I like separating the session’s lifetime from the control connection. Next question: how quickly do higher quality and frame rates push the CPU to its limit?
Overlays, picture-in-picture and existing subtitles
The example browser interface demonstrates webcam picture-in-picture, image overlays, filters and subtitle burn-in. Inputs can come from a webcam, a hosted Cloudflare Stream video over HLS or a Stream Live feed over RTMPS; processed output can go to Stream Live or a WebSocket preview.
Applications can update a PNG overlay while processing continues, allowing on-screen annotations to change during a stream. The hosted-video example renders existing embedded captions into the video image. This is subtitle burn-in, not automatic speech recognition or translation.
Processing continues when the controller disconnects
A Worker handles control requests, while a Durable Object coordinates sessions and relays previews. The Media Engine runs inside a Container, with a Go controller handling requests and FFmpeg currently performing the media processing.
The controlling application can disconnect and later reconnect while the Container keeps processing. Sessions have a maximum duration, so processing eventually stops even without further control requests.
Local Docker, private deployment and the public demo
The code can run locally with Docker or be deployed in your own Cloudflare account. The example app uses an Astro frontend and includes Cloudflare Access tooling. Its private owner deployment allows one active session at a time, and another authenticated user cannot stop or replace someone else’s active session.
The hosted public demo uses a different setup: each verified user has a container identity and can run one active session. Global admission controls and concurrency, media and session limits also apply. The private owner configuration is not a ready-made security model for a public multi-user service.
Explore the demo and code—with a CPU ceiling
The hosted playground is at https://playground.streamline-video.workers.dev. Cloudflare publishes the media engine and application packages at https://github.com/cloudflare/streamline, and the example app, deployment profiles and Access tooling at https://github.com/cloudflare/streamline-demo.
This version processes video on Container CPU, which can become a bottleneck at higher quality or frame rates. Hardware acceleration, computer-vision pipelines, WebRTC/MoQ and native video encoding or decoding in Workers are possible future extensions, not features delivered in this release.
Sources and checked date: October 3, 2026
Comments
0No login needed. Edit or delete your comment from the same browser.
All comments 0
한국어 · English · 日本語No comments yet. Start the conversation.