Boombox

For direct media manipulation (e.g., applying AI effects to video or audio), use the RawData endpoint. This allows Boombox to produce or accept AudioPacket and VideoPacket objects.

There are two primary patterns for raw data interaction:

1. Reading from Boombox: Define a RawData output and iterate over the stream using the boombox.read() generator.
2. Writing to Boombox: Define a RawData input and use the boombox.write(packet) method to feed packets into the stream. This is typically used within a with context manager.

# Pattern 1: Reading raw packets from a stream
boombox = Boombox(
    input=any_input_endpoint,
    output=RawData(video=True, audio=True)
)

for packet in boombox.read():
    process_packet(packet)

# Pattern 2: Writing raw packets to a stream
with Boombox(
    input=RawData(video=True, audio=True),
    output=any_output_endpoint
) as boombox:
    for packet in some_packet_generator():
        boombox.write(packet)

This demo demonstrates how to use boomboxlib to receive a WebRTC stream (e.g., from a browser capturing a webcam feed) and convert it into a live HLS playlist.

Prerequisites

1. Install boomboxlib:

   pip install ../.. 

   (Note: The path above assumes you are running the install from within the example directory relative to the repository root. Adjust the path to point to the actual location of the boomboxlib package.)

2. Hardware: A webcam and optionally a microphone.

Usage Steps

1. Start the server: Run the Python application:

   python main.py

2. Start the WebRTC stream: Open http://localhost:8000/sender.html in your browser and click Connect. You should see your webcam feed on this page.
3. View the HLS stream: Open http://localhost:8000/hls.html and click Play. This page provides access to the converted HLS livestream.

This demo demonstrates Boombox's ability to interact with media streams in Python. It captures a WebRTC stream from a browser webcam, processes it to blur faces, distort voices, and transcribe speech into subtitles, and then streams the processed output back to a browser.

Prerequisites

1. Dependencies: Install the required Python packages using pip:

   pip install -r requirements.txt

2. Hardware: A webcam and a microphone are required.

Execution Steps

1. Start the Python script:

   python anonymization_demo.py

2. Connect the Receiver: Open http://localhost:8000/receiver.html in your browser and click Connect.
3. Connect the Sender: Open http://localhost:8000/sender.html in your browser and click Connect.

Expected Result

• The sender page will display the unmodified webcam feed.
• The receiver page will display the processed stream (blurred face, distorted voice, and subtitles).

# Install dependencies
pip install -r requirements.txt

# Run the demo
python anonymization_demo.py

Boombox allows you to read media from a source (like a URL or an Array of streams) and write it to a destination (like a file or an Array). You can iterate over the stream using the .read() method to access individual packets, such as VideoPacket, allowing for real-time transformations (e.g., using NumPy to flip video frames) before writing them to the output via .write(packet).

Always call .close(wait=True) on the output Boombox instance to ensure all buffers are flushed and the stream is properly finalized.

boombox1 = Boombox(
input="https://raw.githubusercontent.com/membraneframework/static/gh-pages/samples/big-buck-bunny/bun33s.mp4",
output=Array(video=True, audio=True),
)
boombox2 = Boombox(input=Array(video=True, audio=True), output="output.mp4")
for b in boombox1.read():
    if isinstance(b, VideoPacket):
        b.payload = np.flip(b.payload, axis=0)
    boombox2.write(b)
boombox2.close(wait=True)