Hacker News Reader

Comment by bentcorner | Hacker News Reader

FYI, if you have an NVIDIA card on your machine, you can run their voice filtering software: https://www.nvidia.com/en-us/geforce/guides/nvidia-rtx-voice...

It works for non-RTX cards as well with some tweaking (https://arstechnica.com/?post_type=post&p=1670164). Works marvelously for removing background noise - mic hum, typing, cars driving by.

One of the nice things about it is that you can also filter the output on your end - so if you're in a meeting with someone who is sitting outside near a busy road, or they have a roommate who is gaming on a mechanical keyboard, you can filter the noise out without missing something if that person decides to speak up.

View on HN

If anyone wants a demo, this guy tried talking with a massive fan active and banging a hammer on his desk: https://twitter.com/theGunrun/status/1252789873699745792

PulseAudio and Sox can do some of it - https://gist.github.com/tiagoengel/334573f7c3b98c27356297d6b...

You would think this sort of functionality would be implemented in sound cards but I can't think of any significant advancement in audio processing in years.

Aw man, I'm quite often found wanting audio compression - narrowing of the difference in loudness of sound output - but neither Win10 nor Kubuntu seem to be able to do this ootb.

It would be really good to have on a TV too (no more super-loud explosions interspersed with barely audible speech), and I'm amazed that YouTube seemingly doesn't normalise sound levels at all.

As a relatively new Win10 user I find access to mic levels being buried in device manager to be crazy. Also, not having a default output mix, apparently ...

At least games seem to offer a good range of sound settings.

Should be doable in any Linux Distro using pulseaudio (which i think kubuntu does): wiht module-ladspa-sink you can apply arbitrary ladspa plugins to your stream, there are plenty. And i think alsa also hast native compression support, but not sure.

Youtube does some loudness normalization, but it only changes the volume for the entire video, not for specific sections. It also only makes videos quieter, it will never make them louder. You can see this in the "Volume / Normalized" item of the "Stats for nerds" menu.

> As a relatively new Win10 user I find access to mic levels being buried in device manager to be crazy.

Right click on the speaker icon on the bottom right notification area, then click "Open Sound settings"

If you want the legacy sound control panel that you may have been used to from pre-Win10, then from this dialog click "Sound Control Panel" on the right. (Or Win+R "mmsys.cpl")

I've been using Krisp.ai for several months now. It works well.

Happy Krisp.ai user here too. It makes a ton of difference.

I run this on a 1060 w/ patch as well as a 2070 natively and it works amazingly. It's a must-have.

There has to be a non-NVIDIA solution to this doesn't there?

Not at the moment from my research a few weeks ago. There's some sort of lock which only allows it to work on the RTX cores.

Try krisp.ai, works really well too.

Reply on news.ycombinator.com