How to Get Microphone Audio Stream Separately from ARIA Glasses? (Demo Request)

[SunHaoranskywalkers]

Could You Provide a Demo Code for Accessing ARIA Glasses Microphone Audio Stream Separately? Currently, Only Video Stream Demo is Available.

[SeaOtocinclus]

Audio data can be retrieved as follow:
https://facebookresearch.github.io/projectaria_tools/docs/data_utilities/advanced_code_snippets/plotting_sensor_data#audio

If you need to export the audio as a wav file, you can use vrs command line client or our python API.

• Here is how to install VRS Tools:
  https://facebookresearch.github.io/projectaria_tools/docs/data_formats/aria_vrs/aria_vrs_tools_installation
  Then you can do vrs extract-audio my_vrs_file.vrs --to image_folder

• From the Python API you can do:

from projectaria_tools.core.vrs import extract_audio_track
extract_audio_track(vrs_file_path, os.path.join(temp_folder, "audio.wav"))

[SunHaoranskywalkers]

Thank you for your previous response regarding extracting audio from VRS files. However, I believe there was a misunderstanding. I am not looking to extract audio from VRS files. My question is specifically about accessing real-time microphone audio streams directly from the ARIA glasses before the audio is saved into a VRS file.

Here are the key points of my question:

I want to access the real-time audio stream from the ARIA glasses' microphone while the device is active, and before the audio is saved into a VRS file.

I am aware of the method to extract audio from VRS files (e.g., using the Python API or VRS command line tools), but this happens after the audio has been recorded and saved, which does not meet my requirements.

After reviewing the documentation, I found no mention of any API or method to access real-time audio streams. Additionally, the aria/sdk/init.py file does not seem to include any references to audio.

Could you confirm if the Project ARIA SDK currently supports accessing real-time audio streams from the glasses before the audio is saved into VRS files? If not, is this feature planned for future releases?

Thank you for your clarification!

[selcuk-meta]

You can add the on_audio_received callback into the AriaVisualizerStreamingClientObserver class. This class is in the visualizer.py file extracted to the projectaria_client_sdk_samples folder in the Project Aria Client SDK Setup step.

The documentation of on_audio_received is in the API reference page. This function will subscribe to the projectaria_tools.core.sensor_data.AudioDataAndRecord data. Then you can use the Project Aria Tools's "Plot raw sensor data" section shared by @SeaOtocinclus above for visualization if needed.