NVIDIA RTX Voice: Noise cancellation with a bit of AI

Published by

Click here to post a comment for NVIDIA RTX Voice: Noise cancellation with a bit of AI on our message forum
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
Denial:

Few things I want to point out: I watch tons of podcasts/twitch streams that feature 2-12 sometimes more people chatting on microphones - most of these streamers/podcasters, the good ones, already employ various noise cancellation/compression software on the stream to improve the quality. They typically aren't playing games when the podcast/stream is going so the idea that it even uses any GPU % is pretty irrelevant for them. I know a few of them that use Krisp, which is a paid software and as demonstrated this seems to do the job better, for free (as long as you have an RTX card). This cancels noise on outgoing and incoming audio streams. So it's not just improving your $x microphone but in use cases like the one I mentioned above (content creators) it's improving everyone's $2 microphone, including people that don't have RTX cards. You keep saying "get a better mic" but in the video above we see a guy who has a Shure SM7B, a $400 microphone, who not only demonstrates that it has value even with that microphone but says before this released he used various other noise suppressing software/hardware. So clearly even with expensive microphones (arguably the best microphone there is) it's still useful. It was trained via cuDNN and we already know the inference from cuDNN can be ported to various other hardware. Obviously Nvidia is probably not going to offer that because they are essentially selling this as a value-add for RTX owners but I don't see the problem when they are offering as a free plugin.
1st, it is fallacious statement that just because someone does not need full GPU performance, it is OK for everyone. Especially when I specifically talked about those who need all GPU horsepower already! It's same as saying that 4C/4T are perfectly fine for gaming because someone who uses them does not notice problem on his subset of use. Then, you read captions on video incorrectly. RTX does worse job filtering out keyboard even on 100% than Krisp. Same goes for Vacuum where you can hear it even as he moves it as far as he can reach. And when it is closer to mic, it is quite loud. Krisp clearly killed it off completely. As far as voice being clear. Krisp with Vacuum had some minor distortion when he had vacuum cleaner in between his mouth and microphone!!! (Which he did not do for RTX. And which is actually amazing result.) And high pitch pipe shaping of voice was worse with RTX too as it was more noticeable and had higher volume even at time when vacuum was 3 times as far from microphone than in Krisp showcase. Advantage of filtering out incoming audio is good. But it is double edged sword. And lets be honest, moment companies making their own solutions see that competition has desirable feature that they can provide without heavy work, they'll add it too. - - - - So I stand with my view that I rather use HW resource on separate compute device (CPU) that is completely unused, than to use shared resource that I actually need for something else. And that people should get appropriate quality equipment for what they intend to do. It's not unlike those people who pay $2000 for PC and then pirate all games. They are cheaping out on something of value that gets economic damage in return. And they act against their own self interest in long run.
https://forums.guru3d.com/data/avatars/m/220/220626.jpg
So RTX Voice is causing microphone companies economic damage, and is ultimately going to hurt those who use it... but Krisp is fine? If it's flaws you're looking for, not only does it not filter correctly sometimes, but I posted two additional issues above you. Luckily, the software is in beta. Very early beta lol. Don't you think you're being a little dramatic @Fox2232 ?
https://forums.guru3d.com/data/avatars/m/166/166096.jpg
RTX Audio seems to be working great so far for me on Discord etc.. But my biggest issue with it so far is that it prohibits my system from going into sleep or hibernation. When looking at powercfg RTX Audio is clearly to blame: C:\WINDOWS\system32>powercfg /requests DISPLAY: None. SYSTEM: [DRIVER] NVIDIA RTX Voice (ROOT\UNNAMED_DEVICE\0002) An audio stream is currently in use. [DRIVER] USB Audio Device (USB\VID_19F7&PID_0001&MI_00\6&1d4d8455&0&0000) An audio stream is currently in use. [DRIVER] USB Audio Device (USB\VID_0D8C&PID_1316&MI_00\6&2ea0e835&0&0000) An audio stream is currently in use. [DRIVER] Legacy Kernel Caller [DRIVER] Legacy Kernel Caller AWAYMODE: None. EXECUTION: None. PERFBOOST: None. ACTIVELOCKSCREEN: None. I have to manually close the application for my machine to be able to go into sleep or hibernation..
https://forums.guru3d.com/data/avatars/m/273/273678.jpg
working as it should then.
https://forums.guru3d.com/data/avatars/m/238/238382.jpg
[youtube=2gCKOzOxMxU] @Hilbert Hagedoorn