[FEAT]: Allow audio source selection in meeting assistant #3181

New issue

Open

opened 2026-02-28 06:32:50 -05:00 by deekerman · 1 comment

deekerman commented

2026-02-28 06:32:50 -05:00

Owner

Originally created by @weha on GitHub (Feb 19, 2026).

What would you like to see?

The meeting assistant now records from the default output in Windows. I don't see a way to change that other then to actually change the default output.
For example I have a headphone where default output and voice are split so it was recording my windows sounds instead of the actual recording.
So an option to override the recording output would be nice.

Originally created by @weha on GitHub (Feb 19, 2026). ### What would you like to see? The meeting assistant now records from the default output in Windows. I don't see a way to change that other then to actually change the default output. For example I have a headphone where default output and voice are split so it was recording my windows sounds instead of the actual recording. So an option to override the recording output would be nice.

deekerman added the

labels

2026-02-28 06:32:50 -05:00

deekerman commented

2026-02-28 06:32:53 -05:00

Author

Owner

@NWinocur commented on GitHub (Feb 27, 2026):

Came here to suggest this, and glad I searched before submitting it as a duplicate issue.

Given the mixers I've used at my workstation ever since I started streaming, I effectively have multiple separate audio devices I'd want AnythingLLM's meeting assistant to use as input: at a minimum, one to select for my own voice, and one to select for the voices of all the other attendees on a given videocall.

The upside of this is that because they're such distinct channels, it could trivialize how easy it is for a voice model to distinguish between speaker voices. It wouldn't have to guess when I'm the one speaking, it'd have certainty which one's me merely by the fact of that audio coming in on its own separate audio device.

My workaround until this use case is properly supported is to record my meeting using OBS with different sources recorded to different audio channels, drag and drop that MKV file into AnythingLLM after the meeting's done, and have it parse the separate audio channels that way. If I try and run AnythingLLM live, it only gets my voice (because that's the OS' default input device), and hears nothing from other speakers on the call.

@NWinocur commented on GitHub (Feb 27, 2026): Came here to suggest this, and glad I searched before submitting it as a duplicate issue. Given the mixers I've used at my workstation ever since I started streaming, I effectively have multiple separate audio devices I'd want AnythingLLM's meeting assistant to use as input: at a minimum, one to select for my own voice, and one to select for the voices of all the other attendees on a given videocall. The upside of this is that because they're such distinct channels, it could trivialize how easy it is for a voice model to distinguish between speaker voices. It wouldn't have to guess when I'm the one speaking, it'd have certainty which one's me merely by the fact of that audio coming in on its own separate audio device. My workaround until this use case is properly supported is to record my meeting using OBS with different sources recorded to different audio channels, drag and drop that MKV file into AnythingLLM after the meeting's done, and have it parse the separate audio channels that way. If I try and run AnythingLLM live, it only gets my voice (because that's the OS' default input device), and hears nothing from other speakers on the call.

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

starred/anything-llm#3181

No description provided.

Rows
Columns