Send and receive media streams

Updated

How to use Server Gateway to send media streams to Video SDK clients and receive media streams from clients.

This page introduces how to use the Server Gateway to send media streams to the client and receive media streams from the client.

Understand the tech

The following figure shows the basic workflow you implement to send and receive audio and video streams using Server Gateway:

Prerequisites

Before you begin, ensure your development environment meets the necessary hardware and software requirements and that you have downloaded the latest Server Gateway SDK. See Integrate the SDK.

Initialize and connect

Before sending and receiving media streams, complete the following steps:

Initialize the SDK

Create an AgoraService instance and call initialize. The AgoraService object persists as long as the server app is running.

The SDK supports user IDs in both integer and string formats. This page demonstrates use of user IDs in integer format where the character set consists of only digits. To learn more about using string user IDs, see Using a string user ID.

agora_service = AgoraService()

config = AgoraServiceConfig()
config.audio_scenario = AudioScenarioType.AUDIO_SCENARIO_CHORUS
config.appid = sample_options.app_id
agora_service.initialize(config)

Connect to a Video SDK channel

After initializing the SDK, follow these steps to connect to a Video SDK channel.

  1. Call create_rtc_connection to create an RTCConnection object for connecting to the Video SDK channel:

    con_config = RTCConnConfig(
        client_role_type=ClientRoleType.CLIENT_ROLE_BROADCASTER,
        channel_profile=ChannelProfileType.CHANNEL_PROFILE_LIVE_BROADCASTING,
    )
    connection = agora_service.create_rtc_connection(con_config)
    if not connection:
        logger.error("create connection failed")
        return
  2. Call register_observer to listen for connection events. The DYSConnectionObserver class inherits from the IRTCConnectionObserver class.

    # DYSConnectionObserver inherits from the IRTCConnectionObserver class
    conn_observer = DYSConnectionObserver()
    connection.register_observer(conn_observer)
  3. Call connect to connect to a channel:

    ret = connection.connect(sample_options.token, channel_id, uid)
    if ret < 0:
        logger.error(f"connect failed: {ret}")
        return

Send media streams to the client

To send media streams to the client, refer to the following steps:

Create a media stream sender

Use the MediaNodeFactory object to create various types of media stream senders:

  • AudioPcmDataSender: Sends audio data in PCM format.
  • VideoFrameSender: Sends video data in YUV format.
  • AudioEncodedFrameSender: Sends encoded audio data.
  • VideoEncodedImageSender: Sends encoded video data.
  1. Create an MediaNodeFactory object:

    media_node_factory = agora_service.create_media_node_factory()
    if not media_node_factory:
        logger.error("create media node factory failed")
        return
  2. According to your requirements, create an IAudioPcmDataSender object for sending audio in PCM format, an IVideoFrameSender object for sending video in YUV format, an IAudioEncodedFrameSender object for sending encoded audio, or an IVideoEncodedImageSender object for sending encoded video:

    # Create a sender for transmitting audio data in PCM format
    pcm_data_sender = media_node_factory.create_audio_pcm_data_sender()
    if not pcm_data_sender:
        logger.error("Failed to create PCM data sender")
        return
    
    # Create a sender for transmitting encoded audio data
    audio_sender = media_node_factory.create_audio_encoded_frame_sender()
    if not audio_sender:
        logger.error("Failed to create audio sender")
        return
    
    # Create a sender for transmitting video data in YUV format
    video_sender = media_node_factory.create_video_frame_sender()
    if not video_sender:
        logger.error("Failed to create video frame sender")
        return
    
    # Create a sender for transmitting encoded video data
    video_sender = media_node_factory.create_video_encoded_image_sender()
    if not video_sender:
        logger.error("Failed to create video sender")
        return
  3. Create a LocalAudioTrack object and a LocalVideoTrack object, corresponding to the local audio track and local video tracks, for publishing to the channel:

    # Create a custom audio track using a sender for transmitting audio data in PCM format
    audio_track = agora_service.create_custom_audio_track_pcm(pcm_data_sender)
    if not audio_track:
        logger.error("Failed to create audio track")
        return
    
    # Create a custom audio track using a sender for transmitting encoded audio data
    audio_track = agora_service.create_custom_audio_track_encoded(audio_sender, 1)
    if not audio_track:
        logger.error("Failed to create audio track")
        return
    
    # Create a custom video track using a sender for transmitting video data in YUV format
    video_track = agora_service.create_custom_video_track_frame(video_sender)
    if not video_track:
        logger.error("Failed to create video track")
        return
    
    # Create a custom video track using a sender for transmitting encoded video data
    video_track = agora_service.create_custom_video_track_encoded(video_sender, sender_options)
    if not video_track:
        logger.error("Failed to create video track")
        return

Send media streams

  1. To enable media tracks, call setEnabled:

    # Enable the local audio track
    audio_track.set_enabled(1)
    
    # Enable the local video track
    video_track.set_enabled(1)
  2. Call publish_audio and publish_video to publish the local audio and video tracks to the channel:

    # Publish the local audio track
    local_user.publish_audio(audio_track)
    
    # Publish the local video track
    local_user.publish_video(video_track)
  3. Start the audio and video transmission thread:

    async def send(self, sample_options: ExampleOptions, pcm_data_sender, yuv_data_sender):
    # Create and start the audio transmission task
    pcm_task = asyncio.create_task(push_pcm_data_from_file(sample_options.sample_rate, sample_options.num_of_channels, pcm_data_sender, sample_options.audio_file, self._exit))
    
    # Create and start the video transmission task
    yuv_task = asyncio.create_task(push_yuv_data_from_file(sample_options.width, sample_options.height, sample_options.fps, yuv_data_sender, sample_options.video_file, self._exit))
    
    # Wait for the audio transmission task to complete
    await pcm_task
    
    # Wait for the video transmission task to complete
    await yuv_task
    
    logger.info("Transmission finished")
  4. Send audio and video data. Refer to the following example to send PCM audio data and YUV video data:

    # Send PCM audio data
    async def push_pcm_data_from_file(sample_rate, num_of_channels, pcm_data_sender: AudioPcmDataSender, audio_file_path, _exit: Event):
        with open(audio_file_path, "rb") as audio_file:
            pcm_sendinterval = 0.1
            pacer_pcm = Pacer(pcm_sendinterval)
            pcm_count = 0
            send_size = int(sample_rate * num_of_channels * pcm_sendinterval * 2)
            frame_buf = bytearray(send_size)
    
            while not _exit.is_set():
                success = audio_file.readinto(frame_buf)
                if not success:
                    audio_file.seek(0)
                    continue
    
                frame = PcmAudioFrame()
                frame.data = frame_buf
                frame.timestamp = 0
                frame.samples_per_channel = int(sample_rate * pcm_sendinterval)
                frame.bytes_per_sample = 2
                frame.number_of_channels = num_of_channels
                frame.sample_rate = sample_rate
    
                ret = pcm_data_sender.send_audio_pcm_data(frame)
                pcm_count += 1
                logger.info(f"Sent PCM: count={pcm_count}, ret={ret}, send_size={send_size}, interval={pcm_sendinterval}")
    
                await pacer_pcm.apace_interval(0.1)
    
            frame_buf = None
    # Send YUV video data
    async def push_yuv_data_from_file(width, height, fps, video_sender: VideoFrameSender, video_file_path, _exit: Event):
        with open(video_file_path, "rb") as video_file:
            yuv_sendinterval = 1.0 / fps
            pacer_yuv = Pacer(yuv_sendinterval)
            yuv_count = 0
            yuv_len = int(width * height * 3 / 2)
            frame_buf = bytearray(yuv_len)
    
            while not _exit.is_set():
                success = video_file.readinto(frame_buf)
                if not success:
                    video_file.seek(0)
                    continue
    
                frame = ExternalVideoFrame()
                frame.buffer = frame_buf
                frame.type = 1
                frame.format = 1
                frame.stride = width
                frame.height = height
                frame.timestamp = 0
                frame.metadata = "hello metadata"
    
                ret = video_sender.send_video_frame(frame)
                yuv_count += 1
                logger.info("Sent YUV: count=%d, ret=%s", yuv_count, ret)
    
                await pacer_yuv.apace_interval(yuv_sendinterval)
    
            frame_buf = None

Receive media streams from the client

To receive media streams from the client, refer to the following steps:

  1. According to your requirements, call the register_audio_frame_observer and register_video_frame_observer methods to register audio and video data observers and listen for related callbacks. The SDK supports the following data observers:

    • IAudioFrameObserver: Monitors audio data.
    • IVideoFrameObserver: Monitors raw video data.
    • IVideoEncodedImageReceiver: Monitors encoded video data.

    The following code demonstrates registering IAudioFrameObserver and IVideoFrameObserver:

    # Call the set_playback_audio_frame_before_mixing_parameters method before register_audio_frame_observer to set the audio data sampling rate
    local_user.set_playback_audio_frame_before_mixing_parameters(1, 16000)
    
    # The SampleAudioFrameObserver class inherits from IAudioFrameObserver
    audio_frame_observer = SampleAudioFrameObserver()
    local_user.register_audio_frame_observer(audio_frame_observer)
    
    # The SampleVideoFrameObserver class inherits from IVideoFrameObserver
    video_frame_observer = SampleVideoFrameObserver()
    local_user.register_video_frame_observer(video_frame_observer)
  2. When the client starts sending a media stream, the SDK receives the media stream through related callbacks. The following example demonstrates receiving encoded video, YUV-format video, and PCM-format audio:

    # Receive encoded video through the on_encoded_video_frame callback and output it as a file
    class SampleVideoEncodedFrameObserver(IVideoEncodedFrameObserver):
        def on_encoded_video_frame(self, uid, image_buffer, length, video_encoded_frame_info):
            file_path = os.path.join(log_folder, str(uid) + '.h264')
            with open(file_path, 'ab') as f:
                f.write(image_buffer[:length])
            return 1
    
    # Receive YUV format video through the on_frame callback and output it as a file
    class SampleVideoFrameObserver(IVideoFrameObserver):
        def on_frame(self, channel_id, remote_uid, frame:VideoFrame):
            file_path = os.path.join(log_folder, channel_id + "_" + remote_uid + '.yuv')
            y_size = frame.y_stride * frame.height
            uv_size = (frame.u_stride * frame.height) // 2  # Corrected the comment to explain the division
            with open(file_path, 'ab') as f:
                f.write(frame.y_buffer[:y_size])
                f.write(frame.u_buffer[:uv_size])
                f.write(frame.v_buffer[:uv_size])
            return 1
    
    # Receive PCM format audio through the on_playback_audio_frame_before_mixing callback and output it as a file
    class SampleAudioFrameObserver(IAudioFrameObserver):
        def on_playback_audio_frame_before_mixing(self, agora_local_user, channelId, uid, audio_frame:AudioFrame):
            file_path = os.path.join(log_folder, channelId + "_" + uid + '.pcm')
            with open(file_path, "ab") as f:
                f.write(audio_frame.buffer)
            return 1

Stop sending and receiving media streams

When you no longer need to send or receive media streams, follow these steps:

  1. To stop sending media streams:

    • Call unpublish_audio and unpublish_video to stop publishing audio and video tracks.
    • Call set_enabled to disable the local audio and video tracks.
    local_user.unpublish_audio(audio_track)
    local_user.unpublish_video(video_track)
    audio_track.set_enabled(0)
    video_track.set_enabled(0)
  2. To stop receiving media streams, call unregister_audio_frame_observer and unregister_video_frame_observer to unregister the audio and video data observers.

    local_user.unregister_audio_frame_observer(audio_frame_observer)
    local_user.unregister_video_frame_observer(video_frame_observer)

Disconnect and release resources

After completing your media sending and receiving tasks, follow these steps to disconnect from the channel and release resources.

It is important to follow the sequence in the sample code to release resources properly.

  1. Call disconnect to disconnect from the channel.

    ret = connection.disconnect()
    if ret < 0:
        logger.error(f"disconnect failed: {ret}")
        return
  2. Call unregister_observer to unregister the connection event observer.

    connection.unregister_observer()
  3. Release the created objects.

    connection.release()
    agora_service.release()

Test the implementation

To test your project, follow these steps:

  1. Fill in the following parameters:

    • appId: The App ID of your Agora project.
    • channelId: A valid channel name, for example, test_example.
    • userId: Your user ID, for example, 6.
    • audioFile: The storage path of the audio data file, for example, ./test_data/demo.pcm. To facilitate testing, Agora provides test data, that you can download and decompress to the Agora-Python-Server-SDK path.
    • sampleRate: The audio sampling rate, for example, 16000.
    • numOfChannels: The number of audio channels, for example, 1.
    # Assume your Python file path is Desktop/videocall.py
    python Desktop/videocall.py --appId=0a****************************99 --channelId=test_dys --userId=6 --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
  2. Run your Python script from the terminal.

  3. Open the Agora Live Interactive Web Demo, fill in the initialization settings, and simulate a client in the browser. Use the same App ID you specified earlier.

  4. Select Basic Video Calling, fill in the same channel name you specified earlier in the Channel field, and click Create Client and Join Channel in that order.

  5. After successfully joining the channel, you can start streaming. When there is a user streaming in the channel, the Web Demo automatically fills in the user ID of the streaming user in the input box under Step 4: Subscribe & Play. Click the Subscribe & Play button to subscribe to the streaming user.

Due to limitations of the Web Demo, follow these steps based on your use case:

  • To test server-sent streams: First, join the channel on the Web, then run the project on the server to send the streams.

  • To test server-received streams: Run the project on the server to connect to the channel first, then join the channel on the Web and send the streams.

Sample project

Agora provides an open-source server-side sample project for your reference. Download it or view the source code for a more complete example.