Optimizing audio for Dragon Naturally Speaking dictation software

[FrobozzWiz]

Very soon I will be switching to a piece of dictation software called Dragon Naturally Speaking. Towards this goal, I have researched and purchased a Rode NT-USB Mini microphone. Now I would like to optimize the sound for the best possible dictation experience possible (i.e. maximize the accuracy).

What I require is for my voice to sound as clear as possible, while cutting out any non-spoken voice sounds. I don't care if my voice sounds like crap, only that it is very clear with little to no background noise.

Any advice would be greatly appreciated as I am still learning OTS and the world of audio processing and microphones. Thanks in advance!

 

[AaronD]

I'd be surprised if the dictation software didn't have its own recommendations. Follow those, and use their support if it doesn't work.

As an audio guy, I can think of some things that might be helpful in general, but they take some audio experience to get right. How many bands and lecturers have you been a Front-of-House Audio Engineer for?
I would expect that whatever processing the dictation algorithms need, is already included in the software, with automatic adjustment so that the user doesn't screw it up and then complain that it doesn't work. Just feed it a raw mic, and it does the rest.

I don't see what this has to do with OBS.

 

[FrobozzWiz]

During my microphone research I read about one called TableMike. They are very expensive but use some sort of algorithm to boost accuracy to 99%. So my thought was to buy a good but not super-expensive mic and replicate that with OBS.

Sounds like that may be more work than it is worth. I will spend a few hours of training and see how it goes as you recommended. My worry was having to do redo hours of training but knowing me, I am probably just trying to get things too perfect. Thanks AaronD.

 

[AaronD]

During my microphone research I read about one called TableMike. They are very expensive but use some sort of algorithm to boost accuracy to 99%. So my thought was to buy a good but not super-expensive mic and replicate that with OBS.

Sounds like that may be more work than it is worth. I will spend a few hours of training and see how it goes as you recommended. My worry was having to do redo hours of training but knowing me, I am probably just trying to get things too perfect. Thanks AaronD.

I think what the expensive thing really does is called "beamforming". It uses an array of onmidirectional mics, and uses time-of-arrival, phase differences, etc. to figure out where each sound is coming from and only accept from a given direction. That direction might be fixed, or it might change based on where it thinks the primary speaker is at the moment (which could be fooled). You're not going to replicate that at all with OBS; it doesn't even have the tools, regardless of your skill in using them.

Besides, OBS is not an audio app anyway. It's a video app. The token audio stuff is...well...okay for casual use, but if you go much beyond the stereotypical bedroom streamer, you really need an external audio rig, and pass that result through OBS unchanged.

Some people use a physical console that has a USB sound card built-in, or have an explicit analog line input that is fed from the console's analog line outputs. Other people bring everything individually into the production PC and use a DAW to mix the audio (Digital Audio Workstation - essentially a "software sound board" and a ton more). Either way, OBS is not involved at all anymore, except as an unchanged passthrough, and maybe a Noise Gate filter to address some low-level analog noise between a console and a line-in. But that's pretty much *all* that OBS does with audio, in a "serious" rig.

It's not really an audio app at all, so don't think of it that way. It's a video app.

 

[FrobozzWiz]

Thank you for the clarification AaronD. That helps a lot!

Based on your information, I proceeded without altering the audio at all. So far Dragon Naturally Speaking is doing a fantastic job, with few corrections and little training.

Down the road I may look into a software DAW for other projects such as adding audio effects to meetings and presentations & leave OBS for the video side.