A media session is a single abstract representation of a single playing source in the Chrome ecosystem. This could be something like an ARC++ app or a browser tab.
Sessions request audio focus from the media session service and the service maintains a connection to each media session on the platform. This can be used to receive metadata and control media playback in a unified way on Chrome.
To get started the following flags should be enabled:
When audio focus is enabled all the different media sessions will request audio focus. A media session can request three different types of audio focus:
Each WebContents (tab) in Chrome has it's own media session and will request Gain audio focus in most cases. If the media is only a few seconds then it will request the ducking focus type.
In ARC++ each app can make audio focus requests using the Android Audio Manager API. Each of these requests is a media session in Chrome and can be any one of the three above audio focus types.
To reduce the impact of single session audio focus we currently have grouping enabled. By default, all media sessions will have a unique “group id”. All browser media session have the same group id. When a session acquires focus it will not pause any other session with the same group id. Likewise, if playback is resumed automatically all sessions with the same group id will be resumed. This allows all the browser tabs to share audio focus and play concurrently, but still behave as individual media sessions. Ducking is unaffected by group ids.
The active media session is the session that would be controlled if there is a user action (e.g. pressing the play/pause key on a keyboard). Part of the state that the media session exposes is
is_controllable. If this boolean is true then the media session can be controlled by the user. The active media session will be the top most controllable media session.
The media session service can be accessed via Mojo and currently lives in the browser process. Clients can implement AudioFocusObserver and add themselves as an observer using the AudioFocusManager mojo API. The observer will then receive notifications when the active media session changes. This can be used to determine whether there is any current media playback.
The media session service also exposes a MediaController mojo API. This can be used to control the active media session. It will automatically route commands to the active media session, even if the media session changes.
There is also a MediaSession mojo API. This is used for session / service communication and should not be used for control..
There is also a MediaSessionObserver mojo API that clients can implement. They can then add themselves as an observer using the
AddObserver call on the
MediaController api. When an observer is added it will be flushed with all the latest information from the currently active media session.
Again, this will automatically route to the active media session. If that changes then we will flush all methods on the observer with the information from the new active media session.
There is a MockMediaSession C++ class that can be used for simulating a media session in unit tests. The //services/media_session/public/cpp/test directory also contains a number of other useful test utilities.
Questions? - Feel free to reach out to firstname.lastname@example.org.