Calls opens up, Stream clips live broadcasts, Images gets a drop-in widget and face-aware crops
Cloudflare has been positioning Calls, Stream, and Images as the building blocks for live and interactive media experiences. This week, the company is rolling out updates across all three products, with a particular focus on making it easier to get customer data into these systems and on making pricing more transparent for teams that plan to scale.
Cloudflare Calls exits the invite-only phase
Cloudflare Calls is now in open beta and can be activated directly from the dashboard. The company is also announcing the pricing model for when the beta period ends.
Usage will be free until May 15, 2024. After that date, customers with a Calls subscription get the first terabyte each month at no charge, with additional usage billed at $0.05 per real-time gigabyte. Inbound traffic to Cloudflare remains free of charge.
Clip moments from live streams without re-encoding
Cloudflare Stream is introducing Live Instant Clipping, an open beta feature that lets viewers create short clips from ongoing broadcasts. The new Live Clipping API allows clips to be generated on the fly from an active stream without re-encoding the source video.
The feature was designed around a typical workflow for clipping live content: viewers first need a way to review the preceding moments of a broadcast, then select a start time and duration or end time, and finally get a shareable clip quickly. Cloudflare Stream now offers a manifest that previews recent live content in a rolling window. The response includes timing information that lets clients determine the start and end times of a clip relative to the full broadcast.
When a clip is requested, the platform generates a standalone video file for sharing alongside an HLS manifest for embedding in players. Live clips are free to create and do not count toward storage quotas; playback is billed like standard video delivery.
Images: a pre-built upload widget and automatic face cropping
Cloudflare Images is getting two improvements aimed at speeding up application development and improving how images of people are handled.
First, the upload widget is a pre-built, interactive user interface that lets end users upload images directly into a Cloudflare Images account. The widget is intended for applications that rely on user-generated content and can be integrated either by embedding a script into a static HTML page or by installing a package for a JavaScript framework. A ready-made Worker template is available for deployment to connect a frontend application with Cloudflare Images and to authorize uploads through the widget. The widget is currently in closed beta.
Second, automatic face cropping is being added to Cloudflare Images’ dynamic image manipulation features. The existing gravity parameter already uses saliency detection to pick a focal point based on visually interesting pixels. That capability is now being extended with RetinaFace, a machine learning model that identifies human faces in images to inform cropping decisions.
A new zoom parameter can be combined with face cropping to determine how tightly an image should be cropped toward the detected face. This is designed for use cases like generating profile picture thumbnails from user-uploaded photos. Face cropping is also in closed beta.
