I tried to publish local image to video track, and the memory blows up

Thanks for reaching out, Jan. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

I am using Python

await self.room.local_participant.publish_track(self._video_track)

self._video_source.capture_frame(
                            frame=self._image_frame,
                            timestamp_us=int(datetime.datetime.now(datetime.UTC).timestamp())
                            * 1000000,  # Convert to microseconds
                        )

To help you optimize memory usage when publishing a local image to a video track in Python, here are a couple of handy tips: • Efficient Frame Buffer Usage: Try reusing a single frame buffer instead of creating new ones for every frame. For example:

argb_frame = bytearray(WIDTH * HEIGHT * 4)
while True:
    # Update frame content
    frame = rtc.VideoFrame(WIDTH, HEIGHT, rtc.VideoBufferType.RGBA, argb_frame)
    source.capture_frame(frame)

• Proper Resource Cleanup: Make sure to close your video streams properly after you're done to free up memory:

if video_stream:
    await video_stream.aclose()

Following these steps should make managing memory a bit smoother when you're publishing your local image to a video track. If you have any questions or need a hand with anything else, just let me know! Sources: examples/publish_hue.py | Vision Agent Quickstart | LiveKit Docs