Recording support and need help with Opus oggwriter

[moksh-samespace]

Any plan to add participant-wise audio track recording directly within the SFU ?

For now, I’m handling recording externally by tapping into each participant’s track at the source and writing the RTP packets into a file using oggwriter. So for every participant, a separate .ogg file is generated.

But the recorded .ogg files are often shorter in duration than the actual call. For example, in a 5-minute call, the resulting audio file might only be 3 minutes long. This discrepancy is more noticeable when participants are silent or intermittently speaking.

I suspect this could be due to:
• Opus packet timestamp gaps when participants are not transmitting (e.g., silence).
• No silence padding added during recording.
• The oggwriter not compensating for real-time stream duration.

Any Idea how can I solve this?

[tyohan]

This library is focus on SFU related features only so if it is not related to distribute the RTP packets from a publisher to subscribers then I don't think it will add to this library including this track recording.

For your case this can be your options:

• Opus codec use DTX feature that will not transmit any packet on silence and it enabled by default in DefaultClientOptions(). Turn this off may be helpful.
• You might want to explore the WebM recording as well as alternative if oggwriter is not working as expected
• I added recording feature as well to my Inlive API but I decode the Opus to PCM and record it as WAV instead. I have the example in this PR