On-demand and live, usually two stacks
A video library and a live operation have traditionally meant two different systems: one to host and serve on-demand, another to ingest, transcode, time-shift, and restream live, each with its own delivery to wire up and its own bill. Hosted Media Services puts on-demand and live on one backend that signs and delivers every play.
One backend, on-demand and live
HMS does the ingesting, transcoding, recording, signing, and delivery; the console is the control surface on top. Select the backend or a surface to see what it does.
On-demand: the video library
Files arrive through resumable, chunked uploads and appear the moment they land, while transcoding produces streaming renditions, captions, and a storyboard in the background. Playback is a full HLS player on signed, expiring URLs, with storyboard scrubbing, picture-in-picture, chapters, and intro markers.
A video library, from upload to playback
A video enters by upload and is ready to stream once HMS has transcoded it. The library is the workspace for everything that follows: organizing, editing, and playing back what you have uploaded.
- Step 1
Upload
Resumable and chunked. A dropped connection or a closed laptop resumes where it left off, in a floating panel that follows you across the app.
- Step 2
Transcode
Renditions, captions, and a storyboard are generated in the background. A video reads as processing until they are ready, and the player waits for them.
- Step 3
Deliver
A full HLS player streams on signed, expiring URLs, re-signed on demand, with storyboard scrubbing as you seek and picture-in-picture that survives navigation.
Organize
Free-text and shared programmable tags, season and episode numbers, custom thumbnails, intro markers, and chapters. A bulk-edit dialog applies status and tag changes across a whole selection at once.
Find
Search spans title, description, and the streaming key; filters narrow by status. A banner offers a refresh when new content arrives while you are looking at the list.
Play
Storyboard thumbnails preview as you scrub, keyboard shortcuts cover frame-stepping and seeking, and any video pops out into a picture-in-picture player. The detail page pairs the metadata form with the player and its chapter tools.
The library reads your content for you
Two AI layers are native to the library, turning a raw upload into something an editor can curate rather than catalog by hand.
Extractions
On ingest, each video is analyzed into a structured summary: a suggested title, short and long descriptions, detected hosts and guests, subjects, top topics, the spoken language, and a speaker count. These are the raw material an editor curates from.
Chapter generation
For a longer analyzed video, the console turns the caption transcript into a clean set of chapters, each with an English and an Arabic title at the right timecode. Generation runs in the background and drops into the chapter editor, replacing nothing until you confirm.
Deeper AI, editable transcription with speaker diarization and full dubbing into other languages, runs in the Inference studio on these same library items. See AI dubbing.
Live channels, streamed and time-shifted
On the live side, HMS ingests a live signal and transcodes it into a watchable channel. That channel is recorded into a time-shift buffer that lets viewers rewind, pause, and catch up rather than only watching the live edge.
The channel keeps recording at the live edge while the viewer watches an earlier point, then catches back up whenever they choose.
Stream and transcode
Point a live signal at HMS and it transcodes the feed into streaming renditions, the same pipeline that backs the on-demand library.
Time-shift (DVR)
Every live channel is recorded into a buffer. Viewers rewind, pause, and catch up, and opening a channel hands off to the HMS player with full seek and time-shift across the recorded window.
Browse and preview
Channels are surfaced as a browse-and-launch grid with thumbnails that refresh in rotation, and a low-latency channel enlarges and plays live with sound on hover.
Take your live channels to the platforms
Restream is the live side's way out. It re-broadcasts a live source, including the channels you run on HMS, to YouTube, Facebook, Twitch, or any custom RTMP, RTMPS, or SRT endpoint.
- YouTube
- Twitch
- Custom RTMP / SRT
One restream targets one destination. Sending a channel to several platforms is simply several restreams, each with its own live status.
One-step setup
Pick a live source previewed as you choose it, name the restream, choose a platform, and paste the stream key. For a custom target, give the server and stream name, and the console assembles the push URL.
Live status, mirrored
Each running restream shows the source-to-destination flow with a live connection indicator, a ticking duration, and the backend worker’s status.
One destination each
Active and recently stopped restreams are listed separately. One restream targets one destination, and stopping it archives the record.
Signed delivery that can go local
HMS does not only store and transcode; it signs and delivers every play, on-demand and live alike. Delivery is built in, and it can ride the foundation’s OpenCache network to serve from inside local networks.
Signed, expiring URLs
Every play is authorized for a short window and re-signed on demand, across the on-demand library and live channels alike.
Embeds
Live channels and library items embed where your audience already is, carrying the same signed playback.
Local via OpenCache
Delivery can ride the foundation’s OpenCache network, serving playback from inside local networks for on-net latency and transit offload.
OpenCache is the foundation’s neutral, shared cache layer that serves content from inside local networks. See OpenCache.
Everyone working against the same backend
Shared by the project
The library, live channels, and restreams live inside a single project workspace, shared by everyone on that project's team rather than exported between tools.
Gated by access policy
Full read and write for the media library and restreams, browse-only for live channels. The producer, the editor, and the live operator each get the access their role needs.
On-demand and live, one backend
An upload, a live channel, and a restream are different views of the same media infrastructure, not separate products bolted together after the fact.
AI native
Every video is analyzed on the way in, and the same items feed transcription and dubbing in the Inference studio without ever leaving the project.
Built for teams running catalog and live at once
Broadcasters with a catalog and live channels
Host the on-demand library and run live channels with a DVR time-shift buffer, restreamed to the platforms, all from one backend, delivered on signed URLs that can ride OpenCache.
Streaming and live operations
Stream and transcode a live signal into a channel, give viewers a rewindable DVR window, and restream the same feed to YouTube, Facebook, Twitch, and custom endpoints at once, with each broadcast's status mirrored in the console.
Bring your media onto one backend
Hosted Media Services is available to MediaGuard partners. Tell us about your organization and what you produce, and we will take it from there.