Cast: Refactoring Cast API's

Main changes:
1. Adding an IntializeAudio and InitializeVideo to CastSender.
2. Separating FrameInput to audio and video which enables initializing each separately.
3. Changing the CastSender and CastReceiver Create functions to return a scoped_ptr.

These changes better align Cast with the Chromium pipeline.

BUG=346822
R=hclam@chromium.org

Review URL: https://codereview.chromium.org/163553006

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@255954 0039d316-1c4b-4281-b951-d872f2087c98

21 files changed, 358 insertions, 377 deletions

diff --git a/chrome/renderer/media/cast_rtp_stream.cc b/chrome/renderer/media/cast_rtp_stream.cc
index 16e5d1e..3a0fb19 100644
--- a/chrome/renderer/media/cast_rtp_stream.cc
+++ b/chrome/renderer/media/cast_rtp_stream.cc
@@ -155,7 +155,8 @@ class CastVideoSink : public base::SupportsWeakPtr<CastVideoSink>,
   // Attach this sink to MediaStreamTrack. This method call must
   // be made on the render thread. Incoming data can then be
   // passed to media::cast::FrameInput on any thread.
-  void AddToTrack(const scoped_refptr<media::cast::FrameInput>& frame_input) {
+  void AddToTrack(
+      const scoped_refptr<media::cast::VideoFrameInput>& frame_input) {
     DCHECK(render_thread_task_runner_->BelongsToCurrentThread());
 
     frame_input_ = frame_input;
@@ -167,7 +168,7 @@ class CastVideoSink : public base::SupportsWeakPtr<CastVideoSink>,
 
  private:
   blink::WebMediaStreamTrack track_;
-  scoped_refptr<media::cast::FrameInput> frame_input_;
+  scoped_refptr<media::cast::VideoFrameInput> frame_input_;
   bool sink_added_;
   CastRtpStream::ErrorCallback error_callback_;
   scoped_refptr<base::SingleThreadTaskRunner> render_thread_task_runner_;
@@ -301,7 +302,8 @@ class CastAudioSink : public base::SupportsWeakPtr<CastAudioSink>,
   }
 
   // See CastVideoSink for details.
-  void AddToTrack(const scoped_refptr<media::cast::FrameInput>& frame_input) {
+  void AddToTrack(
+      const scoped_refptr<media::cast::AudioFrameInput>& frame_input) {
     DCHECK(render_thread_task_runner_->BelongsToCurrentThread());
     frame_input_ = frame_input;
     if (!sink_added_) {
@@ -316,7 +318,7 @@ class CastAudioSink : public base::SupportsWeakPtr<CastAudioSink>,
 
  private:
   blink::WebMediaStreamTrack track_;
-  scoped_refptr<media::cast::FrameInput> frame_input_;
+  scoped_refptr<media::cast::AudioFrameInput> frame_input_;
   bool sink_added_;
   CastRtpStream::ErrorCallback error_callback_;
   base::WeakPtrFactory<CastAudioSink> weak_factory_;
@@ -348,26 +350,19 @@ CastRtpPayloadParams::CastRtpPayloadParams()
       min_bitrate(0),
       channels(0),
       width(0),
-      height(0) {
-}
+      height(0) {}
 
-CastRtpPayloadParams::~CastRtpPayloadParams() {
-}
+CastRtpPayloadParams::~CastRtpPayloadParams() {}
 
-CastRtpParams::CastRtpParams() {
-}
+CastRtpParams::CastRtpParams() {}
 
-CastRtpParams::~CastRtpParams() {
-}
+CastRtpParams::~CastRtpParams() {}
 
 CastRtpStream::CastRtpStream(const blink::WebMediaStreamTrack& track,
                              const scoped_refptr<CastSession>& session)
-    : track_(track),
-      cast_session_(session),
-      weak_factory_(this) {}
+    : track_(track), cast_session_(session), weak_factory_(this) {}
 
-CastRtpStream::~CastRtpStream() {
-}
+CastRtpStream::~CastRtpStream() {}
 
 std::vector<CastRtpParams> CastRtpStream::GetSupportedParams() {
   if (IsAudio())
@@ -376,9 +371,7 @@ std::vector<CastRtpParams> CastRtpStream::GetSupportedParams() {
     return SupportedVideoParams();
 }
 
-CastRtpParams CastRtpStream::GetParams() {
-  return params_;
-}
+CastRtpParams CastRtpStream::GetParams() { return params_; }
 
 void CastRtpStream::Start(const CastRtpParams& params,
                           const base::Closure& start_callback,
@@ -404,8 +397,7 @@ void CastRtpStream::Start(const CastRtpParams& params,
         params.payload.clock_rate));
     cast_session_->StartAudio(
         config,
-        base::Bind(&CastAudioSink::AddToTrack,
-                   audio_sink_->AsWeakPtr()));
+        base::Bind(&CastAudioSink::AddToTrack, audio_sink_->AsWeakPtr()));
     start_callback.Run();
   } else {
     VideoSenderConfig config;
@@ -420,8 +412,7 @@ void CastRtpStream::Start(const CastRtpParams& params,
                                             weak_factory_.GetWeakPtr()))));
     cast_session_->StartVideo(
         config,
-        base::Bind(&CastVideoSink::AddToTrack,
-                   video_sink_->AsWeakPtr()));
+        base::Bind(&CastVideoSink::AddToTrack, video_sink_->AsWeakPtr()));
     start_callback.Run();
   }
 }
diff --git a/chrome/renderer/media/cast_session.cc b/chrome/renderer/media/cast_session.cc
index e1ed6d5..788225c 100644
--- a/chrome/renderer/media/cast_session.cc
+++ b/chrome/renderer/media/cast_session.cc
@@ -24,7 +24,7 @@ CastSession::~CastSession() {
 }
 
 void CastSession::StartAudio(const media::cast::AudioSenderConfig& config,
-                             const FrameInputAvailableCallback& callback) {
+                             const AudioFrameInputAvailableCallback& callback) {
   DCHECK(content::RenderThread::Get()
              ->GetMessageLoop()
              ->message_loop_proxy()
@@ -39,7 +39,7 @@ void CastSession::StartAudio(const media::cast::AudioSenderConfig& config,
 }
 
 void CastSession::StartVideo(const media::cast::VideoSenderConfig& config,
-                             const FrameInputAvailableCallback& callback) {
+                             const VideoFrameInputAvailableCallback& callback) {
   DCHECK(content::RenderThread::Get()
              ->GetMessageLoop()
              ->message_loop_proxy()
diff --git a/chrome/renderer/media/cast_session.h b/chrome/renderer/media/cast_session.h
index ae7d244..15226bf 100644
--- a/chrome/renderer/media/cast_session.h
+++ b/chrome/renderer/media/cast_session.h
@@ -21,7 +21,8 @@ class MessageLoopProxy;
 namespace media {
 class VideoFrame;
 namespace cast {
-class FrameInput;
+class AudioFrameInput;
+class VideoFrameInput;
 struct AudioSenderConfig;
 struct VideoSenderConfig;
 }  // namespace cast
@@ -38,9 +39,10 @@ class CastSessionDelegate;
 // CastSessionDelegate on the IO thread.
 class CastSession : public base::RefCounted<CastSession> {
  public:
-  typedef
-  base::Callback<void(const scoped_refptr<media::cast::FrameInput>&)>
-      FrameInputAvailableCallback;
+  typedef base::Callback<void(const scoped_refptr<
+      media::cast::AudioFrameInput>&)> AudioFrameInputAvailableCallback;
+  typedef base::Callback<void(const scoped_refptr<
+      media::cast::VideoFrameInput>&)> VideoFrameInputAvailableCallback;
   typedef base::Callback<void(const std::vector<char>&)> SendPacketCallback;
   typedef base::Callback<void(scoped_ptr<std::string>)> EventLogsCallback;
   typedef base::Callback<void(scoped_ptr<base::DictionaryValue>)> StatsCallback;
@@ -53,9 +55,9 @@ class CastSession : public base::RefCounted<CastSession> {
   // media::cast::FrameInput will be given through the callback. The
   // callback will be made on the main thread.
   void StartAudio(const media::cast::AudioSenderConfig& config,
-                  const FrameInputAvailableCallback& callback);
+                  const AudioFrameInputAvailableCallback& callback);
   void StartVideo(const media::cast::VideoSenderConfig& config,
-                  const FrameInputAvailableCallback& callback);
+                  const VideoFrameInputAvailableCallback& callback);
   void StartUDP(const net::IPEndPoint& local_endpoint,
                 const net::IPEndPoint& remote_endpoint);
 
diff --git a/chrome/renderer/media/cast_session_delegate.cc b/chrome/renderer/media/cast_session_delegate.cc
index 11f39185..82e3f1d 100644
--- a/chrome/renderer/media/cast_session_delegate.cc
+++ b/chrome/renderer/media/cast_session_delegate.cc
@@ -43,8 +43,7 @@ const int kMaxAudioEventEntries = kMaxSerializedBytes / 75;
 }  // namespace
 
 CastSessionDelegate::CastSessionDelegate()
-    : transport_configured_(false),
-      io_message_loop_proxy_(
+    : io_message_loop_proxy_(
           content::RenderThread::Get()->GetIOMessageLoopProxy()),
       weak_factory_(this) {
   DCHECK(io_message_loop_proxy_);
@@ -63,56 +62,80 @@ CastSessionDelegate::~CastSessionDelegate() {
   }
 }
 
-void CastSessionDelegate::Initialize(
-    const media::cast::CastLoggingConfig& logging_config) {
-  if (cast_environment_)
-    return;  // Already initialized.
-
-  // CastSender uses the renderer's IO thread as the main thread. This reduces
-  // thread hopping for incoming video frames and outgoing network packets.
-  // There's no need to decode so no thread assigned for decoding.
-  cast_environment_ = new CastEnvironment(
-      scoped_ptr<base::TickClock>(new base::DefaultTickClock()).Pass(),
-      base::MessageLoopProxy::current(),
-      g_cast_threads.Get().GetAudioEncodeMessageLoopProxy(),
-      NULL,
-      g_cast_threads.Get().GetVideoEncodeMessageLoopProxy(),
-      NULL,
-      base::MessageLoopProxy::current(),
-      logging_config);
-}
-
 void CastSessionDelegate::StartAudio(
     const AudioSenderConfig& config,
-    const FrameInputAvailableCallback& callback) {
+    const AudioFrameInputAvailableCallback& callback) {
   DCHECK(io_message_loop_proxy_->BelongsToCurrentThread());
 
-  audio_config_.reset(new AudioSenderConfig(config));
-  video_frame_input_available_callback_ = callback;
-  StartSendingInternal();
+  audio_frame_input_available_callback_ = callback;
+  media::cast::transport::CastTransportAudioConfig transport_config;
+  transport_config.base.ssrc = config.sender_ssrc;
+  transport_config.codec = config.codec;
+  transport_config.base.rtp_config = config.rtp_config;
+  transport_config.frequency = config.frequency;
+  transport_config.channels = config.channels;
+  cast_transport_->InitializeAudio(transport_config);
+  cast_sender_->InitializeAudio(
+      config,
+      base::Bind(&CastSessionDelegate::InitializationResult,
+                 weak_factory_.GetWeakPtr()));
 }
 
 void CastSessionDelegate::StartVideo(
     const VideoSenderConfig& config,
-    const FrameInputAvailableCallback& callback) {
+    const VideoFrameInputAvailableCallback& callback) {
   DCHECK(io_message_loop_proxy_->BelongsToCurrentThread());
-  audio_frame_input_available_callback_ = callback;
+  video_frame_input_available_callback_ = callback;
 
-  video_config_.reset(new VideoSenderConfig(config));
-  StartSendingInternal();
+  media::cast::transport::CastTransportVideoConfig transport_config;
+  transport_config.base.ssrc = config.sender_ssrc;
+  transport_config.codec = config.codec;
+  transport_config.base.rtp_config = config.rtp_config;
+  cast_transport_->InitializeVideo(transport_config);
+  // TODO(mikhal): Pass in a valid GpuVideoAcceleratorFactories to support
+  // hardware video encoding.
+  cast_sender_->InitializeVideo(
+      config,
+      base::Bind(&CastSessionDelegate::InitializationResult,
+                 weak_factory_.GetWeakPtr()),
+      NULL /* GPU*/);
 }
 
 void CastSessionDelegate::StartUDP(const net::IPEndPoint& local_endpoint,
                                    const net::IPEndPoint& remote_endpoint) {
   DCHECK(io_message_loop_proxy_->BelongsToCurrentThread());
-  transport_configured_ = true;
-  local_endpoint_ = local_endpoint;
-  remote_endpoint_ = remote_endpoint;
-  StartSendingInternal();
+
+  // CastSender uses the renderer's IO thread as the main thread. This reduces
+  // thread hopping for incoming video frames and outgoing network packets.
+  // There's no need to decode so no thread assigned for decoding.
+  cast_environment_ = new CastEnvironment(
+      scoped_ptr<base::TickClock>(new base::DefaultTickClock()).Pass(),
+      base::MessageLoopProxy::current(),
+      g_cast_threads.Get().GetAudioEncodeMessageLoopProxy(),
+      NULL,
+      g_cast_threads.Get().GetVideoEncodeMessageLoopProxy(),
+      NULL,
+      base::MessageLoopProxy::current(),
+      media::cast::GetLoggingConfigWithRawEventsAndStatsEnabled());
+
+  // Logging: enable raw events and stats collection.
+  media::cast::CastLoggingConfig logging_config =
+      media::cast::GetLoggingConfigWithRawEventsAndStatsEnabled();
+  // Rationale for using unretained: The callback cannot be called after the
+  // destruction of CastTransportSenderIPC, and they both share the same thread.
+  cast_transport_.reset(new CastTransportSenderIPC(
+      local_endpoint,
+      remote_endpoint,
+      base::Bind(&CastSessionDelegate::StatusNotificationCB,
+                 base::Unretained(this)),
+      logging_config,
+      base::Bind(&CastSessionDelegate::LogRawEvents, base::Unretained(this))));
+
+  cast_sender_ = CastSender::Create(cast_environment_, cast_transport_.get());
+  cast_transport_->SetPacketReceiver(cast_sender_->packet_receiver());
 }
 
-void CastSessionDelegate::ToggleLogging(bool is_audio,
-                                        bool enable) {
+void CastSessionDelegate::ToggleLogging(bool is_audio, bool enable) {
   DCHECK(io_message_loop_proxy_->BelongsToCurrentThread());
   if (enable) {
     if (is_audio) {
@@ -148,11 +171,12 @@ void CastSessionDelegate::ToggleLogging(bool is_audio,
 }
 
 void CastSessionDelegate::GetEventLogsAndReset(
-    bool is_audio, const EventLogsCallback& callback) {
+    bool is_audio,
+    const EventLogsCallback& callback) {
   DCHECK(io_message_loop_proxy_->BelongsToCurrentThread());
 
-  media::cast::EncodingEventSubscriber* subscriber = is_audio ?
-      audio_event_subscriber_.get() : video_event_subscriber_.get();
+  media::cast::EncodingEventSubscriber* subscriber =
+      is_audio ? audio_event_subscriber_.get() : video_event_subscriber_.get();
   if (!subscriber) {
     callback.Run(make_scoped_ptr(new std::string).Pass());
     return;
@@ -202,73 +226,20 @@ void CastSessionDelegate::StatusNotificationCB(
   // TODO(hubbe): Call javascript UDPTransport error function.
 }
 
-void CastSessionDelegate::StartSendingInternal() {
-  DCHECK(io_message_loop_proxy_->BelongsToCurrentThread());
-
-  // No transport, wait.
-  if (!transport_configured_)
-    return;
-
-  // No audio or video, wait.
-  if (!audio_config_ || !video_config_)
-    return;
-
-  // Logging: enable raw events and stats collection.
-  media::cast::CastLoggingConfig logging_config =
-      media::cast::GetLoggingConfigWithRawEventsAndStatsEnabled();
-  Initialize(logging_config);
-
-  // Rationale for using unretained: The callback cannot be called after the
-  // destruction of CastTransportSenderIPC, and they both share the same thread.
-  cast_transport_.reset(new CastTransportSenderIPC(
-      local_endpoint_,
-      remote_endpoint_,
-      base::Bind(&CastSessionDelegate::StatusNotificationCB,
-                 base::Unretained(this)),
-      logging_config,
-      base::Bind(&CastSessionDelegate::LogRawEvents,
-                 base::Unretained(this))));
-
-  // TODO(hubbe): set config.aes_key and config.aes_iv_mask.
-  if (audio_config_) {
-    media::cast::transport::CastTransportAudioConfig config;
-    config.base.ssrc = audio_config_->sender_ssrc;
-    config.codec = audio_config_->codec;
-    config.base.rtp_config = audio_config_->rtp_config;
-    config.frequency = audio_config_->frequency;
-    config.channels = audio_config_->channels;
-    cast_transport_->InitializeAudio(config);
-  }
-  if (video_config_) {
-    media::cast::transport::CastTransportVideoConfig config;
-    config.base.ssrc = video_config_->sender_ssrc;
-    config.codec = video_config_->codec;
-    config.base.rtp_config = video_config_->rtp_config;
-    cast_transport_->InitializeVideo(config);
-  }
-
-  cast_sender_.reset(CastSender::CreateCastSender(
-      cast_environment_,
-      audio_config_.get(),
-      video_config_.get(),
-      NULL,  // GPU.
-      base::Bind(&CastSessionDelegate::InitializationResult,
-                 weak_factory_.GetWeakPtr()),
-      cast_transport_.get()));
-  cast_transport_->SetPacketReceiver(cast_sender_->packet_receiver());
-}
-
 void CastSessionDelegate::InitializationResult(
     media::cast::CastInitializationStatus result) const {
   DCHECK(cast_sender_);
 
   // TODO(pwestin): handle the error codes.
-  if (result == media::cast::STATUS_INITIALIZED) {
+  if (result == media::cast::STATUS_AUDIO_INITIALIZED) {
     if (!audio_frame_input_available_callback_.is_null()) {
-      audio_frame_input_available_callback_.Run(cast_sender_->frame_input());
+      audio_frame_input_available_callback_.Run(
+          cast_sender_->audio_frame_input());
     }
+  } else if (result == media::cast::STATUS_VIDEO_INITIALIZED) {
     if (!video_frame_input_available_callback_.is_null()) {
-      video_frame_input_available_callback_.Run(cast_sender_->frame_input());
+      video_frame_input_available_callback_.Run(
+          cast_sender_->video_frame_input());
     }
   }
 }
diff --git a/chrome/renderer/media/cast_session_delegate.h b/chrome/renderer/media/cast_session_delegate.h
index 07dd5ad..f02dfa3 100644
--- a/chrome/renderer/media/cast_session_delegate.h
+++ b/chrome/renderer/media/cast_session_delegate.h
@@ -44,25 +44,32 @@ class CastTransportSender;
 // thread. All methods are accessible only on the IO thread.
 class CastSessionDelegate {
  public:
-  typedef base::Callback<void(const scoped_refptr<media::cast::FrameInput>&)>
-      FrameInputAvailableCallback;
+  typedef base::Callback<void(const scoped_refptr<
+      media::cast::AudioFrameInput>&)> AudioFrameInputAvailableCallback;
+  typedef base::Callback<void(const scoped_refptr<
+      media::cast::VideoFrameInput>&)> VideoFrameInputAvailableCallback;
   typedef base::Callback<void(scoped_ptr<std::string>)> EventLogsCallback;
   typedef base::Callback<void(scoped_ptr<base::DictionaryValue>)> StatsCallback;
 
   CastSessionDelegate();
   virtual ~CastSessionDelegate();
 
+  // This will start the session by configuring and creating the Cast transport
+  // and the Cast sender.
+  // Must be called before initialization of audio or video.
+  void StartUDP(const net::IPEndPoint& local_endpoint,
+                const net::IPEndPoint& remote_endpoint);
+
   // After calling StartAudio() or StartVideo() encoding of that media will
   // begin as soon as data is delivered to its sink, if the second method is
   // called the first media will be restarted. It is strongly recommended not to
   // deliver any data between calling the two methods.
   // It's OK to call only one of the two methods.
+  // StartUDP must be called before these methods.
   void StartAudio(const media::cast::AudioSenderConfig& config,
-                  const FrameInputAvailableCallback& callback);
+                  const AudioFrameInputAvailableCallback& callback);
   void StartVideo(const media::cast::VideoSenderConfig& config,
-                  const FrameInputAvailableCallback& callback);
-  void StartUDP(const net::IPEndPoint& local_endpoint,
-                const net::IPEndPoint& remote_endpoint);
+                  const VideoFrameInputAvailableCallback& callback);
 
   void ToggleLogging(bool is_audio, bool enable);
   void GetEventLogsAndReset(bool is_audio, const EventLogsCallback& callback);
@@ -75,13 +82,6 @@ class CastSessionDelegate {
   void InitializationResult(media::cast::CastInitializationStatus result) const;
 
  private:
-  // Start encoding threads and initialize the CastEnvironment.
-  void Initialize(const media::cast::CastLoggingConfig& logging_config);
-
-  // Configure CastSender. It is ready to accept audio / video frames after
-  // receiving a successful call to InitializationResult.
-  void StartSendingInternal();
-
   void StatusNotificationCB(
       media::cast::transport::CastTransportStatus status);
 
@@ -93,16 +93,8 @@ class CastSessionDelegate {
   scoped_ptr<media::cast::CastSender> cast_sender_;
   scoped_ptr<media::cast::transport::CastTransportSender> cast_transport_;
 
-  // Configuration for audio and video.
-  scoped_ptr<media::cast::AudioSenderConfig> audio_config_;
-  scoped_ptr<media::cast::VideoSenderConfig> video_config_;
-
-  FrameInputAvailableCallback audio_frame_input_available_callback_;
-  FrameInputAvailableCallback video_frame_input_available_callback_;
-
-  net::IPEndPoint local_endpoint_;
-  net::IPEndPoint remote_endpoint_;
-  bool transport_configured_;
+  AudioFrameInputAvailableCallback audio_frame_input_available_callback_;
+  VideoFrameInputAvailableCallback video_frame_input_available_callback_;
 
   scoped_ptr<media::cast::EncodingEventSubscriber> audio_event_subscriber_;
   scoped_ptr<media::cast::EncodingEventSubscriber> video_event_subscriber_;
diff --git a/media/cast/audio_sender/audio_encoder.cc b/media/cast/audio_sender/audio_encoder.cc
index 2363619..efe7a01 100644
--- a/media/cast/audio_sender/audio_encoder.cc
+++ b/media/cast/audio_sender/audio_encoder.cc
@@ -52,16 +52,16 @@ class AudioEncoder::ImplBase : public base::SupportsWeakPtr<ImplBase> {
         sampling_rate % 100 != 0 ||
         samples_per_10ms_ * num_channels_ >
             transport::EncodedAudioFrame::kMaxNumberOfSamples) {
-      initialization_status_ = STATUS_INVALID_AUDIO_CONFIGURATION;
+      cast_initialization_cb_ = STATUS_INVALID_AUDIO_CONFIGURATION;
     } else {
-      initialization_status_ = STATUS_INITIALIZED;
+      cast_initialization_cb_ = STATUS_AUDIO_INITIALIZED;
     }
   }
 
   virtual ~ImplBase() {}
 
   CastInitializationStatus InitializationResult() const {
-    return initialization_status_;
+    return cast_initialization_cb_;
   }
 
   void LogAudioFrameEvent(uint32 rtp_timestamp,
@@ -156,7 +156,7 @@ class AudioEncoder::ImplBase : public base::SupportsWeakPtr<ImplBase> {
   const int num_channels_;
   const int samples_per_10ms_;
   const FrameEncodedCallback callback_;
-  CastInitializationStatus initialization_status_;
+  CastInitializationStatus cast_initialization_cb_;
 
  private:
   // In the case where a call to EncodeAudio() cannot completely fill the
@@ -192,7 +192,7 @@ class AudioEncoder::OpusImpl : public AudioEncoder::ImplBase {
         encoder_memory_(new uint8[opus_encoder_get_size(num_channels)]),
         opus_encoder_(reinterpret_cast<OpusEncoder*>(encoder_memory_.get())),
         buffer_(new float[num_channels * samples_per_10ms_]) {
-    if (ImplBase::initialization_status_ != STATUS_INITIALIZED) {
+    if (ImplBase::cast_initialization_cb_ != STATUS_AUDIO_INITIALIZED) {
       return;
     }
 
@@ -316,7 +316,6 @@ AudioEncoder::AudioEncoder(
   // Note: It doesn't matter which thread constructs AudioEncoder, just so long
   // as all calls to InsertAudio() are by the same thread.
   insert_thread_checker_.DetachFromThread();
-
   switch (audio_config.codec) {
     case transport::kOpus:
       impl_.reset(new OpusImpl(cast_environment,
@@ -340,6 +339,7 @@ AudioEncoder::AudioEncoder(
 AudioEncoder::~AudioEncoder() {}
 
 CastInitializationStatus AudioEncoder::InitializationResult() const {
+  DCHECK(insert_thread_checker_.CalledOnValidThread());
   if (impl_) {
     return impl_->InitializationResult();
   }
diff --git a/media/cast/audio_sender/audio_sender.cc b/media/cast/audio_sender/audio_sender.cc
index 4ca58ea..cd62f9b 100644
--- a/media/cast/audio_sender/audio_sender.cc
+++ b/media/cast/audio_sender/audio_sender.cc
@@ -113,7 +113,7 @@ AudioSender::AudioSender(scoped_refptr<CastEnvironment> cast_environment,
             audio_config.incoming_feedback_ssrc,
             audio_config.rtcp_c_name),
       timers_initialized_(false),
-      initialization_status_(STATUS_INITIALIZED),
+      cast_initialization_cb_(STATUS_AUDIO_UNINITIALIZED),
       weak_factory_(this) {
   rtcp_.SetCastReceiverEventHistorySize(kReceiverRtcpEventHistorySize);
   if (!audio_config.use_external_encoder) {
@@ -122,7 +122,7 @@ AudioSender::AudioSender(scoped_refptr<CastEnvironment> cast_environment,
                          audio_config,
                          base::Bind(&AudioSender::SendEncodedAudioFrame,
                                     weak_factory_.GetWeakPtr()));
-    initialization_status_ = audio_encoder_->InitializationResult();
+    cast_initialization_cb_ = audio_encoder_->InitializationResult();
   }
 }
 
@@ -141,7 +141,6 @@ void AudioSender::InsertAudio(const AudioBus* audio_bus,
                               const base::Closure& done_callback) {
   DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::MAIN));
   DCHECK(audio_encoder_.get()) << "Invalid internal state";
-
   audio_encoder_->InsertAudio(audio_bus, recorded_time, done_callback);
 }
 
diff --git a/media/cast/audio_sender/audio_sender.h b/media/cast/audio_sender/audio_sender.h
index 0399833..6afa7cf 100644
--- a/media/cast/audio_sender/audio_sender.h
+++ b/media/cast/audio_sender/audio_sender.h
@@ -39,7 +39,7 @@ class AudioSender : public base::NonThreadSafe,
   virtual ~AudioSender();
 
   CastInitializationStatus InitializationResult() const {
-    return initialization_status_;
+    return cast_initialization_cb_;
   }
 
   // The |audio_bus| must be valid until the |done_callback| is called.
@@ -85,7 +85,7 @@ class AudioSender : public base::NonThreadSafe,
   scoped_ptr<LocalRtcpAudioSenderFeedback> rtcp_feedback_;
   Rtcp rtcp_;
   bool timers_initialized_;
-  CastInitializationStatus initialization_status_;
+  CastInitializationStatus cast_initialization_cb_;
 
   DISALLOW_COPY_AND_ASSIGN(AudioSender);
 };
diff --git a/media/cast/cast_config.h b/media/cast/cast_config.h
index 61e9620..21caea0 100644
--- a/media/cast/cast_config.h
+++ b/media/cast/cast_config.h
@@ -134,17 +134,6 @@ struct PcmAudioFrame {
 typedef transport::Packet Packet;
 typedef transport::PacketList PacketList;
 
-enum CastInitializationStatus {
-  STATUS_INITIALIZED,
-  STATUS_INVALID_CAST_ENVIRONMENT,
-  STATUS_INVALID_CRYPTO_CONFIGURATION,
-  STATUS_UNSUPPORTED_AUDIO_CODEC,
-  STATUS_INVALID_AUDIO_CONFIGURATION,
-  STATUS_INVALID_VIDEO_CONFIGURATION,
-  STATUS_GPU_ACCELERATION_NOT_SUPPORTED,
-  STATUS_GPU_ACCELERATION_ERROR,
-};
-
 typedef base::Callback<void(CastInitializationStatus)>
     CastInitializationCallback;
 
diff --git a/media/cast/cast_defines.h b/media/cast/cast_defines.h
index 2e0986f..2ee431e 100644
--- a/media/cast/cast_defines.h
+++ b/media/cast/cast_defines.h
@@ -29,6 +29,20 @@ const int kStartRttMs = 20;
 const int64 kCastMessageUpdateIntervalMs = 33;
 const int64 kNackRepeatIntervalMs = 30;
 
+enum CastInitializationStatus {
+  STATUS_AUDIO_UNINITIALIZED,
+  STATUS_VIDEO_UNINITIALIZED,
+  STATUS_AUDIO_INITIALIZED,
+  STATUS_VIDEO_INITIALIZED,
+  STATUS_INVALID_CAST_ENVIRONMENT,
+  STATUS_INVALID_CRYPTO_CONFIGURATION,
+  STATUS_UNSUPPORTED_AUDIO_CODEC,
+  STATUS_INVALID_AUDIO_CONFIGURATION,
+  STATUS_INVALID_VIDEO_CONFIGURATION,
+  STATUS_GPU_ACCELERATION_NOT_SUPPORTED,
+  STATUS_GPU_ACCELERATION_ERROR,
+};
+
 enum DefaultSettings {
   kDefaultAudioEncoderBitrate = 0,  // This means "auto," and may mean VBR.
   kDefaultAudioSamplingRate = 48000,
diff --git a/media/cast/cast_receiver.h b/media/cast/cast_receiver.h
index ec4f0d3..581a271 100644
--- a/media/cast/cast_receiver.h
+++ b/media/cast/cast_receiver.h
@@ -68,7 +68,7 @@ class FrameReceiver : public base::RefCountedThreadSafe<FrameReceiver> {
 // This Class is thread safe.
 class CastReceiver {
  public:
-  static CastReceiver* CreateCastReceiver(
+  static scoped_ptr<CastReceiver> Create(
       scoped_refptr<CastEnvironment> cast_environment,
       const AudioReceiverConfig& audio_config,
       const VideoReceiverConfig& video_config,
diff --git a/media/cast/cast_receiver_impl.cc b/media/cast/cast_receiver_impl.cc
index f9fc5a6..4561fea 100644
--- a/media/cast/cast_receiver_impl.cc
+++ b/media/cast/cast_receiver_impl.cc
@@ -82,13 +82,13 @@ class LocalFrameReceiver : public FrameReceiver {
   VideoReceiver* video_receiver_;
 };
 
-CastReceiver* CastReceiver::CreateCastReceiver(
+scoped_ptr<CastReceiver> CastReceiver::Create(
     scoped_refptr<CastEnvironment> cast_environment,
     const AudioReceiverConfig& audio_config,
     const VideoReceiverConfig& video_config,
     transport::PacketSender* const packet_sender) {
-  return new CastReceiverImpl(
-      cast_environment, audio_config, video_config, packet_sender);
+  return scoped_ptr<CastReceiver>(new CastReceiverImpl(
+      cast_environment, audio_config, video_config, packet_sender));
 }
 
 CastReceiverImpl::CastReceiverImpl(
diff --git a/media/cast/cast_sender.h b/media/cast/cast_sender.h
index a15e6d3..33b9393 100644
--- a/media/cast/cast_sender.h
+++ b/media/cast/cast_sender.h
@@ -2,11 +2,10 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 //
-// This is the main interface for the cast sender. All configuration are done
-// at creation.
+// This is the main interface for the cast sender.
 //
-// The FrameInput and PacketReciever interfaces should normally be accessed from
-// the IO thread. However they are allowed to be called from any thread.
+// The AudioFrameInput, VideoFrameInput and PacketReciever interfaces should
+// be accessed from the main thread.
 
 #ifndef MEDIA_CAST_CAST_SENDER_H_
 #define MEDIA_CAST_CAST_SENDER_H_
@@ -23,23 +22,32 @@
 
 namespace media {
 class AudioBus;
+class GpuVideoAcceleratorFactories;
 class VideoFrame;
-}
 
-namespace media {
 namespace cast {
+class AudioSender;
+class VideoSender;
 
-// This Class is thread safe.
-class FrameInput : public base::RefCountedThreadSafe<FrameInput> {
+class VideoFrameInput : public base::RefCountedThreadSafe<VideoFrameInput> {
  public:
-  // The video_frame must be valid until the callback is called.
-  // The callback is called from the main cast thread as soon as
-  // the encoder is done with the frame; it does not mean that the encoded frame
-  // has been sent out.
+  // Insert video frames into Cast sender. Frames will be encoded, packetized
+  // and sent to the network.
   virtual void InsertRawVideoFrame(
       const scoped_refptr<media::VideoFrame>& video_frame,
       const base::TimeTicks& capture_time) = 0;
 
+ protected:
+  virtual ~VideoFrameInput() {}
+
+ private:
+  friend class base::RefCountedThreadSafe<VideoFrameInput>;
+};
+
+class AudioFrameInput : public base::RefCountedThreadSafe<AudioFrameInput> {
+ public:
+  // Insert audio frames into Cast sender. Frames will be encoded, packetized
+  // and sent to the network.
   // The |audio_bus| must be valid until the |done_callback| is called.
   // The callback is called from the main cast thread as soon as the encoder is
   // done with |audio_bus|; it does not mean that the encoded data has been
@@ -49,36 +57,44 @@ class FrameInput : public base::RefCountedThreadSafe<FrameInput> {
                            const base::Closure& done_callback) = 0;
 
  protected:
-  virtual ~FrameInput() {}
+  virtual ~AudioFrameInput() {}
 
  private:
-  friend class base::RefCountedThreadSafe<FrameInput>;
+  friend class base::RefCountedThreadSafe<AudioFrameInput>;
 };
 
-// This Class is thread safe.
-// The provided CastTransportSender object will always be called from the main
-// cast thread.
-//  At least one of AudioSenderConfig and VideoSenderConfig have to be provided.
+// The provided CastTransportSender and the CastSender should be called from the
+// main thread.
 class CastSender {
  public:
-  static CastSender* CreateCastSender(
+  static scoped_ptr<CastSender> Create(
       scoped_refptr<CastEnvironment> cast_environment,
-      const AudioSenderConfig* audio_config,
-      const VideoSenderConfig* video_config,
-      const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-      const CastInitializationCallback& cast_initialization,
       transport::CastTransportSender* const transport_sender);
 
   virtual ~CastSender() {}
 
-  // All audio and video frames for the session should be inserted to this
-  // object.
-  // Can be called from any thread.
-  virtual scoped_refptr<FrameInput> frame_input() = 0;
+  // All video frames for the session should be inserted to this object.
+  virtual scoped_refptr<VideoFrameInput> video_frame_input() = 0;
+
+  // All audio frames for the session should be inserted to this object.
+  virtual scoped_refptr<AudioFrameInput> audio_frame_input() = 0;
 
   // All RTCP packets for the session should be inserted to this object.
-  // Can be called from any thread.
+  // This function and the callback must be called on the main thread.
   virtual transport::PacketReceiverCallback packet_receiver() = 0;
+
+  // Initialize the audio stack. Must be called in order to send audio frames.
+  // Status of the initialization will be returned on cast_initialization_cb.
+  virtual void InitializeAudio(
+      const AudioSenderConfig& audio_config,
+      const CastInitializationCallback& cast_initialization_cb) = 0;
+
+  // Initialize the video stack. Must be called in order to send video frames.
+  // Status of the initialization will be returned on cast_initialization_cb.
+  virtual void InitializeVideo(
+      const VideoSenderConfig& video_config,
+      const CastInitializationCallback& cast_initialization_cb,
+      const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories) = 0;
 };
 
 }  // namespace cast
diff --git a/media/cast/cast_sender_impl.cc b/media/cast/cast_sender_impl.cc
index 1a97ae4..c461d35 100644
--- a/media/cast/cast_sender_impl.cc
+++ b/media/cast/cast_sender_impl.cc
@@ -12,17 +12,13 @@
 namespace media {
 namespace cast {
 
-// The LocalFrameInput class posts all incoming frames; audio and video to the
-// main cast thread for processing.
-// This make the cast sender interface thread safe.
-class LocalFrameInput : public FrameInput {
+// The LocalVideoFrameInput class posts all incoming video frames to the main
+// cast thread for processing.
+class LocalVideoFrameInput : public VideoFrameInput {
  public:
-  LocalFrameInput(scoped_refptr<CastEnvironment> cast_environment,
-                  base::WeakPtr<AudioSender> audio_sender,
-                  base::WeakPtr<VideoSender> video_sender)
-      : cast_environment_(cast_environment),
-        audio_sender_(audio_sender),
-        video_sender_(video_sender) {}
+  LocalVideoFrameInput(scoped_refptr<CastEnvironment> cast_environment,
+                       base::WeakPtr<VideoSender> video_sender)
+      : cast_environment_(cast_environment), video_sender_(video_sender) {}
 
   virtual void InsertRawVideoFrame(
       const scoped_refptr<media::VideoFrame>& video_frame,
@@ -35,6 +31,26 @@ class LocalFrameInput : public FrameInput {
                                            capture_time));
   }
 
+ protected:
+  virtual ~LocalVideoFrameInput() {}
+
+ private:
+  friend class base::RefCountedThreadSafe<LocalVideoFrameInput>;
+
+  scoped_refptr<CastEnvironment> cast_environment_;
+  base::WeakPtr<VideoSender> video_sender_;
+
+  DISALLOW_COPY_AND_ASSIGN(LocalVideoFrameInput);
+};
+
+// The LocalAudioFrameInput class posts all incoming audio frames to the main
+// cast thread for processing. Therefore frames can be inserted from any thread.
+class LocalAudioFrameInput : public AudioFrameInput {
+ public:
+  LocalAudioFrameInput(scoped_refptr<CastEnvironment> cast_environment,
+                       base::WeakPtr<AudioSender> audio_sender)
+      : cast_environment_(cast_environment), audio_sender_(audio_sender) {}
+
   virtual void InsertAudio(const AudioBus* audio_bus,
                            const base::TimeTicks& recorded_time,
                            const base::Closure& done_callback) OVERRIDE {
@@ -48,96 +64,71 @@ class LocalFrameInput : public FrameInput {
   }
 
  protected:
-  virtual ~LocalFrameInput() {}
+  virtual ~LocalAudioFrameInput() {}
 
  private:
-  friend class base::RefCountedThreadSafe<LocalFrameInput>;
+  friend class base::RefCountedThreadSafe<LocalAudioFrameInput>;
 
   scoped_refptr<CastEnvironment> cast_environment_;
   base::WeakPtr<AudioSender> audio_sender_;
-  base::WeakPtr<VideoSender> video_sender_;
 
-  DISALLOW_COPY_AND_ASSIGN(LocalFrameInput);
+  DISALLOW_COPY_AND_ASSIGN(LocalAudioFrameInput);
 };
 
-CastSender* CastSender::CreateCastSender(
+scoped_ptr<CastSender> CastSender::Create(
     scoped_refptr<CastEnvironment> cast_environment,
-    const AudioSenderConfig* audio_config,
-    const VideoSenderConfig* video_config,
-    const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-    const CastInitializationCallback& initialization_status,
     transport::CastTransportSender* const transport_sender) {
   CHECK(cast_environment);
-  return new CastSenderImpl(cast_environment,
-                            audio_config,
-                            video_config,
-                            gpu_factories,
-                            initialization_status,
-                            transport_sender);
+  return scoped_ptr<CastSender>(
+      new CastSenderImpl(cast_environment, transport_sender));
 }
 
 CastSenderImpl::CastSenderImpl(
     scoped_refptr<CastEnvironment> cast_environment,
-    const AudioSenderConfig* audio_config,
-    const VideoSenderConfig* video_config,
-    const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-    const CastInitializationCallback& initialization_status,
     transport::CastTransportSender* const transport_sender)
-    : initialization_callback_(initialization_status),
-      packet_receiver_(
-          base::Bind(&CastSenderImpl::ReceivedPacket, base::Unretained(this))),
-      cast_environment_(cast_environment),
+    : cast_environment_(cast_environment),
+      transport_sender_(transport_sender),
       weak_factory_(this) {
   CHECK(cast_environment);
-  CHECK(audio_config || video_config);
-
-  base::WeakPtr<AudioSender> audio_sender_ptr;
-  base::WeakPtr<VideoSender> video_sender_ptr;
-
-  if (audio_config) {
-    CHECK(audio_config->use_external_encoder ||
-          cast_environment->HasAudioEncoderThread());
-
-    audio_sender_.reset(
-        new AudioSender(cast_environment, *audio_config, transport_sender));
-    ssrc_of_audio_sender_ = audio_config->incoming_feedback_ssrc;
-    audio_sender_ptr = audio_sender_->AsWeakPtr();
-
-    CastInitializationStatus status = audio_sender_->InitializationResult();
-    if (status != STATUS_INITIALIZED || !video_config) {
-      if (status == STATUS_INITIALIZED && !video_config) {
-        // Audio only.
-        frame_input_ = new LocalFrameInput(
-            cast_environment, audio_sender_ptr, video_sender_ptr);
-      }
-      cast_environment->PostTask(
-          CastEnvironment::MAIN,
-          FROM_HERE,
-          base::Bind(&CastSenderImpl::InitializationResult,
-                     weak_factory_.GetWeakPtr(),
-                     status));
-      return;
-    }
-  }
-  if (video_config) {
-    CHECK(video_config->use_external_encoder ||
-          cast_environment->HasVideoEncoderThread());
-
-    video_sender_.reset(
-        new VideoSender(cast_environment,
-                        *video_config,
-                        gpu_factories,
-                        base::Bind(&CastSenderImpl::InitializationResult,
-                                   weak_factory_.GetWeakPtr()),
-                        transport_sender));
-    video_sender_ptr = video_sender_->AsWeakPtr();
-    ssrc_of_video_sender_ = video_config->incoming_feedback_ssrc;
+}
+
+void CastSenderImpl::InitializeAudio(
+    const AudioSenderConfig& audio_config,
+    const CastInitializationCallback& cast_initialization_cb) {
+  DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::MAIN));
+  CHECK(audio_config.use_external_encoder ||
+        cast_environment_->HasAudioEncoderThread());
+
+  audio_sender_.reset(
+      new AudioSender(cast_environment_, audio_config, transport_sender_));
+
+  CastInitializationStatus status = audio_sender_->InitializationResult();
+
+  if (status == STATUS_AUDIO_INITIALIZED) {
+    ssrc_of_audio_sender_ = audio_config.incoming_feedback_ssrc;
+    audio_frame_input_ =
+        new LocalAudioFrameInput(cast_environment_, audio_sender_->AsWeakPtr());
   }
-  frame_input_ =
-      new LocalFrameInput(cast_environment, audio_sender_ptr, video_sender_ptr);
+  cast_initialization_cb.Run(status);
+}
 
-  // Handing over responsibility to call NotifyInitialization to the
-  // video sender.
+void CastSenderImpl::InitializeVideo(
+    const VideoSenderConfig& video_config,
+    const CastInitializationCallback& cast_initialization_cb,
+    const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories) {
+  DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::MAIN));
+  CHECK(video_config.use_external_encoder ||
+        cast_environment_->HasVideoEncoderThread());
+
+  video_sender_.reset(new VideoSender(cast_environment_,
+                                      video_config,
+                                      gpu_factories,
+                                      cast_initialization_cb,
+                                      transport_sender_));
+
+  ssrc_of_video_sender_ = video_config.incoming_feedback_ssrc;
+  video_frame_input_ =
+      new LocalVideoFrameInput(cast_environment_, video_sender_->AsWeakPtr());
 }
 
 CastSenderImpl::~CastSenderImpl() {}
@@ -203,16 +194,17 @@ void CastSenderImpl::ReceivedPacket(scoped_ptr<Packet> packet) {
   }
 }
 
-scoped_refptr<FrameInput> CastSenderImpl::frame_input() { return frame_input_; }
+scoped_refptr<AudioFrameInput> CastSenderImpl::audio_frame_input() {
+  return audio_frame_input_;
+}
 
-transport::PacketReceiverCallback CastSenderImpl::packet_receiver() {
-  return packet_receiver_;
-  return base::Bind(&CastSenderImpl::ReceivedPacket, base::Unretained(this));
+scoped_refptr<VideoFrameInput> CastSenderImpl::video_frame_input() {
+  return video_frame_input_;
 }
 
-void CastSenderImpl::InitializationResult(CastInitializationStatus status)
-    const {
-  initialization_callback_.Run(status);
+transport::PacketReceiverCallback CastSenderImpl::packet_receiver() {
+  return base::Bind(&CastSenderImpl::ReceivedPacket,
+                    weak_factory_.GetWeakPtr());
 }
 
 }  // namespace cast
diff --git a/media/cast/cast_sender_impl.h b/media/cast/cast_sender_impl.h
index bfb4041..85b8669 100644
--- a/media/cast/cast_sender_impl.h
+++ b/media/cast/cast_sender_impl.h
@@ -8,49 +8,53 @@
 #include "base/memory/scoped_ptr.h"
 #include "media/cast/audio_sender/audio_sender.h"
 #include "media/cast/cast_config.h"
+#include "media/cast/cast_defines.h"
 #include "media/cast/cast_environment.h"
 #include "media/cast/cast_sender.h"
 #include "media/cast/video_sender/video_sender.h"
 
 namespace media {
 class VideoFrame;
-}
 
-namespace media {
 namespace cast {
-
 class AudioSender;
 class VideoSender;
 
-// This calls is a pure owner class that group all required sending objects
-// together such as pacer, packet receiver, frame input, audio and video sender.
+// This class combines all required sending objects such as the audio and video
+// senders, pacer, packet receiver and frame input.
 class CastSenderImpl : public CastSender {
  public:
-  CastSenderImpl(
-      scoped_refptr<CastEnvironment> cast_environment,
-      const AudioSenderConfig* audio_config,
-      const VideoSenderConfig* video_config,
-      const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-      const CastInitializationCallback& initialization_status,
-      transport::CastTransportSender* const transport_sender);
+  CastSenderImpl(scoped_refptr<CastEnvironment> cast_environment,
+                 transport::CastTransportSender* const transport_sender);
+
+  virtual void InitializeAudio(
+      const AudioSenderConfig& audio_config,
+      const CastInitializationCallback& cast_initialization_cb) OVERRIDE;
+  virtual void InitializeVideo(
+      const VideoSenderConfig& video_config,
+      const CastInitializationCallback& cast_initialization_cb,
+      const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories)
+      OVERRIDE;
 
   virtual ~CastSenderImpl();
 
-  virtual scoped_refptr<FrameInput> frame_input() OVERRIDE;
+  virtual scoped_refptr<AudioFrameInput> audio_frame_input() OVERRIDE;
+  virtual scoped_refptr<VideoFrameInput> video_frame_input() OVERRIDE;
+
   virtual transport::PacketReceiverCallback packet_receiver() OVERRIDE;
 
  private:
   void ReceivedPacket(scoped_ptr<Packet> packet);
-  // Used to trampoline the result back on the correct thread. And guaranteed
-  // not to be called until the creation is complete.
-  void InitializationResult(CastInitializationStatus status) const;
 
   CastInitializationCallback initialization_callback_;
   scoped_ptr<AudioSender> audio_sender_;
   scoped_ptr<VideoSender> video_sender_;
-  scoped_refptr<FrameInput> frame_input_;
-  transport::PacketReceiverCallback packet_receiver_;
+  scoped_refptr<AudioFrameInput> audio_frame_input_;
+  scoped_refptr<VideoFrameInput> video_frame_input_;
   scoped_refptr<CastEnvironment> cast_environment_;
+  // The transport sender is owned by the owner of the CastSender, and should be
+  // valid throughout the lifetime of the CastSender.
+  transport::CastTransportSender* const transport_sender_;
   uint32 ssrc_of_audio_sender_;
   uint32 ssrc_of_video_sender_;
   base::WeakPtrFactory<CastSenderImpl> weak_factory_;
diff --git a/media/cast/test/end2end_unittest.cc b/media/cast/test/end2end_unittest.cc
index 1524c8f..71b2c3e 100644
--- a/media/cast/test/end2end_unittest.cc
+++ b/media/cast/test/end2end_unittest.cc
@@ -95,6 +95,14 @@ void UpdateCastTransportStatus(transport::CastTransportStatus status) {
   EXPECT_TRUE(result);
 }
 
+void AudioInitializationStatus(CastInitializationStatus status) {
+  EXPECT_EQ(STATUS_AUDIO_INITIALIZED, status);
+}
+
+void VideoInitializationStatus(CastInitializationStatus status) {
+  EXPECT_EQ(STATUS_VIDEO_INITIALIZED, status);
+}
+
 // This is wrapped in a struct because it needs to be put into a std::map.
 typedef struct {
   int counter[kNumOfLoggingEvents];
@@ -396,8 +404,8 @@ class End2EndTest : public ::testing::Test {
       : start_time_(),
         testing_clock_sender_(new base::SimpleTestTickClock()),
         testing_clock_receiver_(new base::SimpleTestTickClock()),
-        task_runner_(new test::FakeSingleThreadTaskRunner(
-            testing_clock_sender_)),
+        task_runner_(
+            new test::FakeSingleThreadTaskRunner(testing_clock_sender_)),
         logging_config_(GetLoggingConfigWithRawEventsAndStatsEnabled()),
         cast_environment_sender_(new CastEnvironment(
             scoped_ptr<base::TickClock>(testing_clock_sender_).Pass(),
@@ -492,11 +500,10 @@ class End2EndTest : public ::testing::Test {
   }
 
   void Create() {
-    cast_receiver_.reset(
-        CastReceiver::CreateCastReceiver(cast_environment_receiver_,
-                                         audio_receiver_config_,
-                                         video_receiver_config_,
-                                         &receiver_to_sender_));
+    cast_receiver_ = CastReceiver::Create(cast_environment_receiver_,
+                                          audio_receiver_config_,
+                                          video_receiver_config_,
+                                          &receiver_to_sender_);
     net::IPEndPoint dummy_endpoint;
     transport_sender_.reset(new transport::CastTransportSenderImpl(
         NULL,
@@ -512,18 +519,21 @@ class End2EndTest : public ::testing::Test {
     transport_sender_->InitializeAudio(transport_audio_config_);
     transport_sender_->InitializeVideo(transport_video_config_);
 
-    cast_sender_.reset(CastSender::CreateCastSender(
-        cast_environment_sender_,
-        &audio_sender_config_,
-        &video_sender_config_,
-        NULL,
-        base::Bind(&End2EndTest::InitializationResult, base::Unretained(this)),
-        transport_sender_.get()));
+    cast_sender_ =
+        CastSender::Create(cast_environment_sender_, transport_sender_.get());
+
+    // Initializing audio and video senders.
+    cast_sender_->InitializeAudio(audio_sender_config_,
+                                  base::Bind(&AudioInitializationStatus));
+    cast_sender_->InitializeVideo(
+        video_sender_config_, base::Bind(&VideoInitializationStatus), NULL);
 
     receiver_to_sender_.SetPacketReceiver(cast_sender_->packet_receiver());
     sender_to_receiver_.SetPacketReceiver(cast_receiver_->packet_receiver());
 
-    frame_input_ = cast_sender_->frame_input();
+    audio_frame_input_ = cast_sender_->audio_frame_input();
+    video_frame_input_ = cast_sender_->video_frame_input();
+
     frame_receiver_ = cast_receiver_->frame_receiver();
 
     audio_bus_factory_.reset(
@@ -555,7 +565,7 @@ class End2EndTest : public ::testing::Test {
         media::VideoFrame::CreateFrame(
             VideoFrame::I420, size, gfx::Rect(size), size, time_diff);
     PopulateVideoFrame(video_frame, start_value);
-    frame_input_->InsertRawVideoFrame(video_frame, capture_time);
+    video_frame_input_->InsertRawVideoFrame(video_frame, capture_time);
   }
 
   void RunTasks(int during_ms) {
@@ -567,10 +577,6 @@ class End2EndTest : public ::testing::Test {
     }
   }
 
-  void InitializationResult(CastInitializationStatus result) {
-    EXPECT_EQ(result, STATUS_INITIALIZED);
-  }
-
   void LogRawEvents(const std::vector<PacketEvent>& packet_events) {
     EXPECT_FALSE(packet_events.empty());
     for (std::vector<media::cast::PacketEvent>::const_iterator it =
@@ -608,7 +614,8 @@ class End2EndTest : public ::testing::Test {
 
   scoped_ptr<CastReceiver> cast_receiver_;
   scoped_ptr<CastSender> cast_sender_;
-  scoped_refptr<FrameInput> frame_input_;
+  scoped_refptr<AudioFrameInput> audio_frame_input_;
+  scoped_refptr<VideoFrameInput> video_frame_input_;
   scoped_refptr<FrameReceiver> frame_receiver_;
 
   scoped_refptr<TestReceiverAudioCallback> test_receiver_audio_callback_;
@@ -620,7 +627,6 @@ class End2EndTest : public ::testing::Test {
   std::vector<FrameEvent> frame_events_;
   std::vector<PacketEvent> packet_events_;
   std::vector<GenericEvent> generic_events_;
-
   // |transport_sender_| has a RepeatingTimer which needs a MessageLoop.
   base::MessageLoop message_loop_;
 };
@@ -657,7 +663,7 @@ TEST_F(End2EndTest, LoopNoLossPcm16) {
     }
 
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -714,7 +720,7 @@ TEST_F(End2EndTest, LoopNoLossPcm16ExternalDecoder) {
         send_time);
 
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -749,7 +755,7 @@ TEST_F(End2EndTest, LoopNoLossOpus) {
     }
 
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -799,7 +805,7 @@ TEST_F(End2EndTest, StartSenderBeforeReceiver) {
         base::TimeDelta::FromMilliseconds(10) * num_10ms_blocks));
 
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -841,7 +847,7 @@ TEST_F(End2EndTest, StartSenderBeforeReceiver) {
     }
 
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -1072,7 +1078,7 @@ TEST_F(End2EndTest, CryptoAudio) {
           send_time);
     }
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -1250,7 +1256,7 @@ TEST_F(End2EndTest, AudioLogging) {
     }
 
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         send_time,
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
diff --git a/media/cast/test/sender.cc b/media/cast/test/sender.cc
index 7a0ff7e..90d48c9 100644
--- a/media/cast/test/sender.cc
+++ b/media/cast/test/sender.cc
@@ -205,11 +205,13 @@ class SendProcess {
   SendProcess(scoped_refptr<base::SingleThreadTaskRunner> thread_proxy,
               base::TickClock* clock,
               const VideoSenderConfig& video_config,
-              FrameInput* frame_input)
+              scoped_refptr<AudioFrameInput> audio_frame_input,
+              scoped_refptr<VideoFrameInput> video_frame_input)
       : test_app_thread_proxy_(thread_proxy),
         video_config_(video_config),
         audio_diff_(kFrameTimerMs),
-        frame_input_(frame_input),
+        audio_frame_input_(audio_frame_input),
+        video_frame_input_(video_frame_input),
         synthetic_count_(0),
         clock_(clock),
         start_time_(),
@@ -245,7 +247,7 @@ class SendProcess {
     scoped_ptr<AudioBus> audio_bus(audio_bus_factory_->NextAudioBus(
         base::TimeDelta::FromMilliseconds(10) * num_10ms_blocks));
     AudioBus* const audio_bus_ptr = audio_bus.get();
-    frame_input_->InsertAudio(
+    audio_frame_input_->InsertAudio(
         audio_bus_ptr,
         clock_->NowTicks(),
         base::Bind(&OwnThatAudioBus, base::Passed(&audio_bus)));
@@ -277,21 +279,21 @@ class SendProcess {
       test_app_thread_proxy_->PostDelayedTask(
           FROM_HERE,
           base::Bind(&SendProcess::SendVideoFrameOnTime,
-                     base::Unretained(this),
+                     weak_factory_.GetWeakPtr(),
                      video_frame),
           video_frame_time - elapsed_time);
     } else {
       test_app_thread_proxy_->PostTask(
           FROM_HERE,
           base::Bind(&SendProcess::SendVideoFrameOnTime,
-                     base::Unretained(this),
+                     weak_factory_.GetWeakPtr(),
                      video_frame));
     }
   }
 
   void SendVideoFrameOnTime(scoped_refptr<media::VideoFrame> video_frame) {
     send_time_ = clock_->NowTicks();
-    frame_input_->InsertRawVideoFrame(video_frame, send_time_);
+    video_frame_input_->InsertRawVideoFrame(video_frame, send_time_);
     test_app_thread_proxy_->PostTask(
         FROM_HERE, base::Bind(&SendProcess::SendFrame, base::Unretained(this)));
   }
@@ -300,7 +302,8 @@ class SendProcess {
   scoped_refptr<base::SingleThreadTaskRunner> test_app_thread_proxy_;
   const VideoSenderConfig video_config_;
   int audio_diff_;
-  const scoped_refptr<FrameInput> frame_input_;
+  const scoped_refptr<AudioFrameInput> audio_frame_input_;
+  const scoped_refptr<VideoFrameInput> video_frame_input_;
   FILE* video_file_;
   uint8 synthetic_count_;
   base::TickClock* const clock_;  // Not owned by this class.
@@ -336,8 +339,9 @@ void LogRawEvents(
 }
 
 void InitializationResult(media::cast::CastInitializationStatus result) {
-  CHECK_EQ(result, media::cast::STATUS_INITIALIZED);
-  VLOG(1) << "Cast Sender initialized";
+  bool end_result = result == media::cast::STATUS_AUDIO_INITIALIZED ||
+                    result == media::cast::STATUS_VIDEO_INITIALIZED;
+  CHECK(end_result) << "Cast sender uninitialized";
 }
 
 net::IPEndPoint CreateUDPAddress(std::string ip_str, int port) {
@@ -359,14 +363,14 @@ void WriteLogsToFileAndStopSubscribing(
   media::cast::FrameEventMap frame_events;
   media::cast::PacketEventMap packet_events;
   media::cast::RtpTimestamp first_rtp_timestamp;
-  video_event_subscriber->GetEventsAndReset(&frame_events, &packet_events,
-                                            &first_rtp_timestamp);
+  video_event_subscriber->GetEventsAndReset(
+      &frame_events, &packet_events, &first_rtp_timestamp);
 
   VLOG(0) << "Video frame map size: " << frame_events.size();
   VLOG(0) << "Video packet map size: " << packet_events.size();
 
-  if (!serializer.SerializeEventsForStream(false, frame_events, packet_events,
-                                           first_rtp_timestamp)) {
+  if (!serializer.SerializeEventsForStream(
+          false, frame_events, packet_events, first_rtp_timestamp)) {
     VLOG(1) << "Failed to serialize video events.";
     return;
   }
@@ -377,14 +381,14 @@ void WriteLogsToFileAndStopSubscribing(
   // Serialize audio events.
   cast_environment->Logging()->RemoveRawEventSubscriber(
       audio_event_subscriber.get());
-  audio_event_subscriber->GetEventsAndReset(&frame_events, &packet_events,
-                                            &first_rtp_timestamp);
+  audio_event_subscriber->GetEventsAndReset(
+      &frame_events, &packet_events, &first_rtp_timestamp);
 
   VLOG(0) << "Audio frame map size: " << frame_events.size();
   VLOG(0) << "Audio packet map size: " << packet_events.size();
 
-  if (!serializer.SerializeEventsForStream(true, frame_events, packet_events,
-                                           first_rtp_timestamp)) {
+  if (!serializer.SerializeEventsForStream(
+          true, frame_events, packet_events, first_rtp_timestamp)) {
     VLOG(1) << "Failed to serialize audio events.";
     return;
   }
@@ -406,7 +410,6 @@ void WriteLogsToFileAndStopSubscribing(
 
 int main(int argc, char** argv) {
   base::AtExitManager at_exit;
-  VLOG(1) << "Cast Sender";
   base::Thread test_thread("Cast sender test app thread");
   base::Thread audio_thread("Cast audio encoder thread");
   base::Thread video_thread("Cast video encoder thread");
@@ -472,23 +475,25 @@ int main(int argc, char** argv) {
   transport_sender->InitializeAudio(transport_audio_config);
   transport_sender->InitializeVideo(transport_video_config);
 
-  scoped_ptr<media::cast::CastSender> cast_sender(
-      media::cast::CastSender::CreateCastSender(
-          cast_environment,
-          &audio_config,
-          &video_config,
-          NULL,  // gpu_factories.
-          base::Bind(&InitializationResult),
-          transport_sender.get()));
+  scoped_ptr<media::cast::CastSender> cast_sender =
+      media::cast::CastSender::Create(cast_environment, transport_sender.get());
+
+  cast_sender->InitializeVideo(
+      video_config, base::Bind(&InitializationResult), NULL);
+  cast_sender->InitializeAudio(audio_config, base::Bind(&InitializationResult));
 
   transport_sender->SetPacketReceiver(cast_sender->packet_receiver());
 
-  media::cast::FrameInput* frame_input = cast_sender->frame_input();
+  scoped_refptr<media::cast::AudioFrameInput> audio_frame_input =
+      cast_sender->audio_frame_input();
+  scoped_refptr<media::cast::VideoFrameInput> video_frame_input =
+      cast_sender->video_frame_input();
   scoped_ptr<media::cast::SendProcess> send_process(
       new media::cast::SendProcess(test_thread.message_loop_proxy(),
                                    cast_environment->Clock(),
                                    video_config,
-                                   frame_input));
+                                   audio_frame_input,
+                                   video_frame_input));
 
   // Set up event subscribers.
   int logging_duration = media::cast::GetLoggingDuration();
diff --git a/media/cast/test/utility/in_process_receiver.cc b/media/cast/test/utility/in_process_receiver.cc
index f916dbb..c8dfc28 100644
--- a/media/cast/test/utility/in_process_receiver.cc
+++ b/media/cast/test/utility/in_process_receiver.cc
@@ -67,8 +67,8 @@ void InProcessReceiver::StartOnMainThread() {
                        remote_end_point_,
                        base::Bind(&InProcessReceiver::UpdateCastTransportStatus,
                                   base::Unretained(this))));
-  cast_receiver_.reset(CastReceiver::CreateCastReceiver(
-      cast_environment_, audio_config_, video_config_, transport_.get()));
+  cast_receiver_ = CastReceiver::Create(
+      cast_environment_, audio_config_, video_config_, transport_.get());
 
   // TODO(hubbe): Make the cast receiver do this automatically.
   transport_->StartReceiving(cast_receiver_->packet_receiver());
diff --git a/media/cast/video_sender/video_sender.cc b/media/cast/video_sender/video_sender.cc
index 148086e..1582bc8 100644
--- a/media/cast/video_sender/video_sender.cc
+++ b/media/cast/video_sender/video_sender.cc
@@ -95,7 +95,7 @@ VideoSender::VideoSender(
     scoped_refptr<CastEnvironment> cast_environment,
     const VideoSenderConfig& video_config,
     const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-    const CastInitializationCallback& initialization_status,
+    const CastInitializationCallback& cast_initialization_cb,
     transport::CastTransportSender* const transport_sender)
     : rtp_max_delay_(base::TimeDelta::FromMilliseconds(
           video_config.rtp_config.max_delay_ms)),
@@ -149,12 +149,12 @@ VideoSender::VideoSender(
                video_config.rtcp_c_name));
   rtcp_->SetCastReceiverEventHistorySize(kReceiverRtcpEventHistorySize);
 
-  // TODO(pwestin): pass cast_initialization to |video_encoder_|
+  // TODO(pwestin): pass cast_initialization_cb to |video_encoder_|
   // and remove this call.
   cast_environment_->PostTask(
       CastEnvironment::MAIN,
       FROM_HERE,
-      base::Bind(initialization_status, STATUS_INITIALIZED));
+      base::Bind(cast_initialization_cb, STATUS_VIDEO_INITIALIZED));
   cast_environment_->Logging()->AddRawEventSubscriber(&event_subscriber_);
 
   memset(frame_id_to_rtp_timestamp_, 0, sizeof(frame_id_to_rtp_timestamp_));
diff --git a/media/cast/video_sender/video_sender.h b/media/cast/video_sender/video_sender.h
index 8a59707..8c20a40 100644
--- a/media/cast/video_sender/video_sender.h
+++ b/media/cast/video_sender/video_sender.h
@@ -47,7 +47,7 @@ class VideoSender : public base::NonThreadSafe,
   VideoSender(scoped_refptr<CastEnvironment> cast_environment,
               const VideoSenderConfig& video_config,
               const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-              const CastInitializationCallback& initialization_status,
+              const CastInitializationCallback& cast_initialization_cb,
               transport::CastTransportSender* const transport_sender);
 
   virtual ~VideoSender();
diff --git a/media/cast/video_sender/video_sender_unittest.cc b/media/cast/video_sender/video_sender_unittest.cc
index 982f7de..0bd4f2f 100644
--- a/media/cast/video_sender/video_sender_unittest.cc
+++ b/media/cast/video_sender/video_sender_unittest.cc
@@ -63,12 +63,12 @@ class PeerVideoSender : public VideoSender {
       scoped_refptr<CastEnvironment> cast_environment,
       const VideoSenderConfig& video_config,
       const scoped_refptr<GpuVideoAcceleratorFactories>& gpu_factories,
-      const CastInitializationCallback& initialization_status,
+      const CastInitializationCallback& cast_initialization_cb,
       transport::CastTransportSender* const transport_sender)
       : VideoSender(cast_environment,
                     video_config,
                     gpu_factories,
-                    initialization_status,
+                    cast_initialization_cb,
                     transport_sender) {}
   using VideoSender::OnReceivedCastFeedback;
 };
@@ -174,7 +174,7 @@ class VideoSenderTest : public ::testing::Test {
   }
 
   void InitializationResult(CastInitializationStatus result) {
-    EXPECT_EQ(result, STATUS_INITIALIZED);
+    EXPECT_EQ(result, STATUS_VIDEO_INITIALIZED);
   }
 
   base::SimpleTestTickClock* testing_clock_;  // Owned by CastEnvironment.