Stream Module

This module contains streaming functionality.

RTMP Streaming

RTMP streaming module for FMUS-VID.

This module provides classes for capturing from RTMP streams, outputting to RTMP streams, and monitoring stream health.

class fmusvid.stream.rtmp.RTMPCapture(rtmp_url, width=None, height=None, fps=None, audio_sample_rate=16000, buffer_size=10, gpu_acceleration=True)

Capture frames and audio from an RTMP stream.

This class uses FFmpeg to read from an RTMP stream and provides methods to access frames and audio data.

__init__(rtmp_url, width=None, height=None, fps=None, audio_sample_rate=16000, buffer_size=10, gpu_acceleration=True)

Initialize RTMP capture.

Parameters:

• rtmp_url (str) – RTMP URL to capture from (e.g., ‘rtmp://localhost/live/stream’)

• width (Optional[int]) – Output width (None to use source width)

• height (Optional[int]) – Output height (None to use source height)

• fps (Optional[float]) – Output frame rate (None to use source frame rate)

• audio_sample_rate (int) – Audio sample rate in Hz

• buffer_size (int) – Size of frame buffer

• gpu_acceleration (bool) – Whether to use GPU acceleration if available

async start()

Start capturing from the RTMP stream.

async stop()

Stop capturing from the RTMP stream.

async get_frame()

Get the next video frame from the stream.

Returns:

RGB image as numpy array

Return type:

np.ndarray

Raises:

RuntimeError – If the stream is not running

async get_audio(duration=0.1)

Get audio samples for a specified duration.

Parameters:

duration (float) – Duration of audio to get in seconds

Returns:

Audio samples as numpy array

Return type:

np.ndarray

async get_frame_and_audio()

Get the next video frame and corresponding audio.

Returns:

RGB image and audio samples

Return type:

Tuple[np.ndarray, np.ndarray]

async aiter_frames()

Async iterator for video frames.

Yields:

np.ndarray – RGB image as numpy array

Return type:

AsyncGenerator[ndarray, None]

async aiter_frames_with_audio()

Async iterator for video frames with corresponding audio.

Yields:

Tuple[np.ndarray, np.ndarray] – RGB image and audio samples

Return type:

AsyncGenerator[Tuple[ndarray, ndarray], None]

iter_frames()

Iterator for video frames (non-async version).

Yields:

np.ndarray – RGB image as numpy array

Return type:

Generator[ndarray, None, None]

iter_frames_with_audio()

Iterator for video frames with corresponding audio (non-async version).

Yields:

Tuple[np.ndarray, np.ndarray] – RGB image and audio samples

Return type:

Generator[Tuple[ndarray, ndarray], None, None]

__del__()

Clean up resources when the object is garbage collected.

class fmusvid.stream.rtmp.RTMPOutput(rtmp_url, width, height, fps=30.0, bitrate='2500k', format='flv', vcodec='libx264', acodec='aac', audio_sample_rate=44100, preset='ultrafast', gpu_acceleration=True)

Output frames to an RTMP stream.

This class uses FFmpeg to send frames to an RTMP stream.

__init__(rtmp_url, width, height, fps=30.0, bitrate='2500k', format='flv', vcodec='libx264', acodec='aac', audio_sample_rate=44100, preset='ultrafast', gpu_acceleration=True)

Initialize RTMP output.

Parameters:

• rtmp_url (str) – RTMP URL to output to (e.g., ‘rtmp://localhost/live/stream’)

• width (int) – Width of output video

• height (int) – Height of output video

• fps (float) – Frame rate of output video

• bitrate (str) – Video bitrate

• format (str) – Output format (usually ‘flv’ for RTMP)

• vcodec (str) – Video codec

• acodec (str) – Audio codec

• audio_sample_rate (int) – Audio sample rate in Hz

• preset (str) – FFmpeg preset for encoding speed vs quality tradeoff

• gpu_acceleration (bool) – Whether to use GPU acceleration if available

async start()

Start the RTMP output stream.

async stop()

Stop the RTMP output stream.

async write_frame(frame)

Write a video frame to the RTMP stream.

Parameters:

frame (ndarray) – RGB image as numpy array

Raises:

RuntimeError – If the stream is not running

async write_frame_with_audio(frame, audio)

Write a video frame with audio to the RTMP stream.

Note: This is a simplified implementation. In practice, handling audio in this way is complex and would require more sophisticated audio buffering and synchronization.

Parameters:

• frame (ndarray) – RGB image as numpy array

• audio (ndarray) – Audio samples as numpy array

async write_frame_async(frame)

Alias for write_frame to maintain naming consistency with other async methods.

Parameters:

frame (ndarray) – RGB image as numpy array

async close()

Alias for stop().

__del__()

Clean up resources when the object is garbage collected.

class fmusvid.stream.rtmp.StreamMonitor(metrics=None)

Monitor health and performance of video streams.

This class provides metrics and alerting for stream health.

__init__(metrics=None)

Initialize stream monitor.

Parameters:

metrics (Optional[List[str]]) – List of metrics to monitor (e.g., [“bitrate”, “fps”, “latency”])

attach_stream(stream, name=None)

Attach a stream to monitor.

Parameters:

• stream (Union[RTMPCapture, RTMPOutput]) – RTMP stream to monitor

• name (Optional[str]) – Optional name for the stream

detach_stream(name)

Detach a stream from monitoring.

Parameters:

name (str) – Name of the stream to detach

get_metric(metric, stream_name=None)

Get the current value of a metric.

Parameters:

• metric (str) – Name of the metric

• stream_name (Optional[str]) – Name of the stream (None for average across all streams)

Returns:

Current value of the metric

Return type:

float

get_alerts(stream_name=None)

Get active alerts for streams.

Parameters:

stream_name (Optional[str]) – Name of the stream (None for all streams)

Returns:

List of alert dictionaries

Return type:

List[Dict]

__del__()

Clean up resources when the object is garbage collected.