Fingerprinting primary speaker

[patchenm]

This feature is about dealing / ignoring background noise when the background noise is another speaker.

What I've found is that Deepgram seems to do a good job of isolating the primary speakers voice and ignoring all the background speaking while the primary speaker is talking.

However, if the primary speaker pauses to take a breath or think and someone is talking in the background, then the background voices immediately start getting transcribed.

I have another discussion on whether there is any work-arounds available right now, but to the extent not, of all potential feature requests, the ability to quickly 'fingerprint' the primary speaker so that if they pause for a second, background voices aren't picked up, would be a huge benefit.

The other issue is that if you're in a room with other people talking, even if those voices are filtered out while you're talking, once you're done, deepgram will immediately start transcribing them, so if you're waiting for deepgram to 'catch up' with final results vs interm before pressing continue, you can end up in a situation where that opportunity doesn't present itself because deepgram starts picking up the other voices and thinks the user isn't finished talking.