Downloading media

This documentation may be out-of-date. Please refer to the documentation for the latest ExoPlayer release on

ExoPlayer provides functionality to download media for offline playback. In most use cases it’s desirable for downloads to continue even when your app is in the background. For these use cases your app should subclass DownloadService, and send commands to the service to add, remove and control the downloads. The diagram below shows the main classes that are involved.

for downloading media. The arrow directions indicate the flow of data.
Figure 1. Classes for downloading media. The arrow directions indicate the flow of data.
  • DownloadService: Wraps a DownloadManager and forwards commands to it. The service allows the DownloadManager to keep running even when the app is in the background.
  • DownloadManager: Manages multiple downloads, loading (and storing) their states from (and to) a DownloadIndex, starting and stopping downloads based on requirements such as network connectivity, and so on. To download the content, the manager will typically read the data being downloaded from a HttpDataSource, and write it into a Cache.
  • DownloadIndex: Persists the states of the downloads.

Creating a DownloadService

To create a DownloadService, you need to subclass it and implement its abstract methods:

  • getDownloadManager(): Returns the DownloadManager to be used.
  • getScheduler(): Returns an optional Scheduler, which can restart the service when requirements needed for pending downloads to progress are met. ExoPlayer provides these implementations:
  • getForegroundNotification(): Returns a notification to be displayed when the service is running in the foreground. You can use DownloadNotificationHelper.buildProgressNotification to create a notification in default style.

Finally, you need to define the service in your AndroidManifest.xml file:

<service android:name="com.myapp.MyDownloadService"
  <!-- This is needed for Scheduler -->
    <action android:name=""/>
    <category android:name="android.intent.category.DEFAULT"/>

See DemoDownloadService and AndroidManifest.xml in the ExoPlayer demo app for a concrete example.

Creating a DownloadManager

The following code snippet demonstrates how to instantiate a DownloadManager, which can be returned by getDownloadManager() in your DownloadService:

// Note: This should be a singleton in your app.
databaseProvider = new StandaloneDatabaseProvider(context);

// A download cache should not evict media, so should use a NoopCacheEvictor.
downloadCache = new SimpleCache(
    new NoOpCacheEvictor(),

// Create a factory for reading the data from the network.
dataSourceFactory = new DefaultHttpDataSource.Factory();

// Choose an executor for downloading data. Using Runnable::run will cause each download task to
// download data on its own thread. Passing an executor that uses multiple threads will speed up
// download tasks that can be split into smaller parts for parallel execution. Applications that
// already have an executor for background downloads may wish to reuse their existing executor.
Executor downloadExecutor = Runnable::run;

// Create the download manager.
downloadManager = new DownloadManager(

// Optionally, setters can be called to configure the download manager.

See DemoUtil in the demo app for a concrete example.

Adding a download

To add a download you need to create a DownloadRequest and send it to your DownloadService. For adaptive streams DownloadHelper can be used to help build a DownloadRequest, as described further down this page. The example below shows how to create a download request:

DownloadRequest downloadRequest =
    new DownloadRequest.Builder(contentId, contentUri).build();

where contentId is a unique identifier for the content. In simple cases, the contentUri can often be used as the contentId, however apps are free to use whatever ID scheme best suits their use case. DownloadRequest.Builder also has some optional setters. For example, setKeySetId and setData can be used to set DRM and custom data that the app wishes to associate with the download, respectively. The content’s MIME type can also be specified using setMimeType, as a hint for cases where the content type cannot be inferred from contentUri.

Once created, the request can be sent to the DownloadService to add the download:

    /* foreground= */ false)

where MyDownloadService is the app’s DownloadService subclass, and the foreground parameter controls whether the service will be started in the foreground. If your app is already in the foreground then the foreground parameter should normally be set to false, since the DownloadService will put itself in the foreground if it determines that it has work to do.

Removing downloads

A download can be removed by sending a remove command to the DownloadService, where contentId identifies the download to be removed:

    /* foreground= */ false)

You can also remove all downloaded data with DownloadService.sendRemoveAllDownloads.

Starting and stopping downloads

A download will only progress if four conditions are met:

  • The download doesn’t have a stop reason.
  • Downloads aren’t paused.
  • The requirements for downloads to progress are met. Requirements can specify constraints on the allowed network types, as well as whether the device should be idle or connected to a charger.
  • The maximum number of parallel downloads is not exceeded.

All of these conditions can be controlled by sending commands to your DownloadService.

Setting and clearing download stop reasons

It’s possible to set a reason for one or all downloads being stopped:

// Set the stop reason for a single download.
    /* foreground= */ false);

// Clear the stop reason for a single download.
    /* foreground= */ false);

where stopReason can be any non-zero value (Download.STOP_REASON_NONE = 0 is a special value meaning that the download is not stopped). Apps that have multiple reasons for stopping downloads can use different values to keep track of why each download is stopped. Setting and clearing the stop reason for all downloads works the same way as setting and clearing the stop reason for a single download, except that contentId should be set to null.

Setting a stop reason does not remove a download. The partial download will be retained, and clearing the stop reason will cause the download to continue.

When a download has a non-zero stop reason, it will be in the Download.STATE_STOPPED state. Stop reasons are persisted in the DownloadIndex, and so are retained if the application process is killed and later restarted.

Pausing and resuming all downloads

All downloads can be paused and resumed as follows:

// Pause all downloads.
    /* foreground= */ false);

// Resume all downloads.
    /* foreground= */ false);

When downloads are paused, they will be in the Download.STATE_QUEUED state. Unlike setting stop reasons, this approach does not persist any state changes. It only affects the runtime state of the DownloadManager.

Setting the requirements for downloads to progress

Requirements can be used to specify constraints that must be met for downloads to proceed. The requirements can be set by calling DownloadManager.setRequirements() when creating the DownloadManager, as in the example above. They can also be changed dynamically by sending a command to the DownloadService:

// Set the download requirements.
    /* foreground= */ false);

When a download cannot proceed because the requirements are not met, it will be in the Download.STATE_QUEUED state. You can query the not met requirements with DownloadManager.getNotMetRequirements().

Setting the maximum number of parallel downloads

The maximum number of parallel downloads can be set by calling DownloadManager.setMaxParallelDownloads(). This would normally be done when creating the DownloadManager, as in the example above.

When a download cannot proceed because the maximum number of parallel downloads are already in progress, it will be in the Download.STATE_QUEUED state.

Querying downloads

The DownloadIndex of a DownloadManager can be queried for the state of all downloads, including those that have completed or failed. The DownloadIndex can be obtained by calling DownloadManager.getDownloadIndex(). A cursor that iterates over all downloads can then be obtained by calling DownloadIndex.getDownloads(). Alternatively, the state of a single download can be queried by calling DownloadIndex.getDownload().

DownloadManager also provides DownloadManager.getCurrentDownloads(), which returns the state of current (i.e. not completed or failed) downloads only. This method is useful for updating notifications and other UI components that display the progress and status of current downloads.

Listening to downloads

You can add a listener to DownloadManager to be informed when current downloads change state:

    new DownloadManager.Listener() {
      // Override methods of interest here.

See DownloadManagerListener in the demo app’s DownloadTracker class for a concrete example.

Download progress updates do not trigger calls on DownloadManager.Listener. To update a UI component that shows download progress, you should periodically query the DownloadManager at your desired update rate. DownloadService contains an example of this, which periodically updates the service foreground notification.

Playing downloaded content

Playing downloaded content is similar to playing online content, except that data is read from the download Cache instead of over the network.

It’s important that you do not try and read files directly from the download directory. Instead, use ExoPlayer library classes as described below.

To play downloaded content, create a CacheDataSource.Factory using the same Cache instance that was used for downloading, and inject it into DefaultMediaSourceFactory when building the player:

// Create a read-only cache data source factory using the download cache.
DataSource.Factory cacheDataSourceFactory =
    new CacheDataSource.Factory()
        .setCacheWriteDataSinkFactory(null); // Disable writing.

ExoPlayer player = new ExoPlayer.Builder(context)
        new DefaultMediaSourceFactory(context)

If the same player instance will also be used to play non-downloaded content then the CacheDataSource.Factory should be configured as read-only to avoid downloading that content as well during playback.

Once the player has been configured with the CacheDataSource.Factory, it will have access to the downloaded content for playback. Playing a download is then as simple as passing the corresponding MediaItem to the player. A MediaItem can be obtained from a Download using Download.request.toMediaItem, or directly from a DownloadRequest using DownloadRequest.toMediaItem.

MediaSource configuration

The example above makes the download cache available for playback of all MediaItems. It’s also possible to make the download cache available for individual MediaSource instances, which can be passed directly to the player:

ProgressiveMediaSource mediaSource =
    new ProgressiveMediaSource.Factory(cacheDataSourceFactory)

Downloading and playing adaptive streams

Adaptive streams (e.g. DASH, SmoothStreaming and HLS) normally contain multiple media tracks. There are often multiple tracks that contain the same content in different qualities (e.g. SD, HD and 4K video tracks). There may also be multiple tracks of the same type containing different content (e.g. multiple audio tracks in different languages).

For streaming playbacks, a track selector can be used to choose which of the tracks are played. Similarly, for downloading, a DownloadHelper can be used to choose which of the tracks are downloaded. Typical usage of a DownloadHelper follows these steps:

  1. Build a DownloadHelper using one of the DownloadHelper.forMediaItem methods. Prepare the helper and wait for the callback.
    DownloadHelper downloadHelper =
            new DefaultRenderersFactory(context),
  2. Optionally, inspect the default selected tracks using getMappedTrackInfo and getTrackSelections, and make adjustments using clearTrackSelections, replaceTrackSelections and addTrackSelection.
  3. Create a DownloadRequest for the selected tracks by calling getDownloadRequest. The request can be passed to your DownloadService to add the download, as described above.
  4. Release the helper using release().

Playback of downloaded adaptive content requires configuring the player and passing the corresponding MediaItem, as described above.

When building the MediaItem, MediaItem.playbackProperties.streamKeys must be set to match those in the DownloadRequest so that the player only tries to play the subset of tracks that have been downloaded. Using Download.request.toMediaItem and DownloadRequest.toMediaItem to build the MediaItem will take care of this for you.

If you see data being requested from the network when trying to play downloaded adaptive content, the most likely cause is that the player is trying to adapt to a track that was not downloaded. Ensure you’ve set the stream keys correctly.