Like there’s video AI to enhance video, but podcasts have so many bad connections with guests it’s horrible to listen to. They need to run out through an audio AI.
(Don’t even get me started on asking the President questions when the fucking helicopter is in the background. Who tf can hear anything?)
Actually they’ve been here for a very long time.
DeBird is the oldest commercial ML/Audio product I know. It’s around since 2017.
Accentize plugins are the commercial ones.
Also there’s a Waves one.
Nvidia’s RTX Voice can do it in real time since 2020.
And lastly on the open-source end, there’s Intel’s OpenVino for Audacity.
RTX Voice was renamed to Nvidia Broadcast: https://www.nvidia.com/en-us/geforce/broadcasting/broadcast-app/
Davinci resolve studio has had a pretty great AI de-noisier for awhile now! They also use AI to be able to turn off certain instruments/parts of songs ( and its local!)
Got a good recommendation for the video one you mentioned? I have some videos digitized from VHS tapes that I’d like to clean up. (I’m only interested in models I can run locally, BTW, for obvious privacy reasons.)
Depends on what kind of ai and how much control you want over it. There are some consumer stuff like adobe or elevenlabs, but if a professional is using ai it’s likely either waves or izotope as they give way more control. Personally I use waves, their clarity pro plugins are amazing, but I know several people who use izotope as well.
The problem is doing it in real time and having it sound halfway decent. The only companies who have managed that is waves and Yamaha, the later of which requires hardware that costs $70k+.
I’m using https://auphonic.com/ for my podcasts.
If you are fine with non-free shit that will probably train with your audio, try out adobe’s AI.
Not aware of any but have a look at https://replicate.com/explore
