How to extract audio from mp4

This approach won't work for chatbot, as it's deployed in an Azure Web App. We can't rely on installing any ex locally, nor can we save files to a specific folder in the web app's environment.

What we need is a solution that doesn't depend on any local installation or file system storage.

Ideally, we should be able to extract the audio from the video (MP4) directly into a memory stream (e.g., as MP3 or WAV), and then use that stream to send the audio to Azure Speech Service for transcription completely in memory, without creating physical files.

In Teams personal chats, there’s no direct option to send an audio message. But Teams does have a “Record video clip” feature, where users can just speak and send a short video (which is actually an MP4 file).

Now if we want to use that voice for speech-to-text (like with Azure Speech Service), we’ll need to extract the audio from the MP4 first.

Best way is to convert that MP4 into an MP3 or WAV file. For Azure, WAV is better — ideally mono channel, 16kHz sample rate.

You can use ffmpeg to do that:

bash

ffmpeg -i input.mp4 -ac 1 -ar 16000 output.wav

Once you’ve got the .wav file, just send it to Azure Speech API and get the transcription.

So yeah, even though Teams doesn’t support direct audio messages in personal chats, this is a pretty solid workaround. Users just record a quick video, we extract the audio, and we’re good to go.