Many of my videos involve people speaking in a noisy environment, so I have been testing AI voice isolation. Here is a test video, I would be interested what you guys think, and any other suggestions you have for better sound.
I'm using a Sennheiser MKE440 here which is stereo. I like the spacial atmosphere although of course the speaker is mono. It is pretty good front/back if I face away from the noise source, but in this case it is directed at the machinery too. I sometimes use a wireless lapel mic but that is omni-directional so picks up nearby noise wherever it is.
I am impressed at the way AI can totally remove unwanted background noise. So much so that it sounds unnatural without a little bit, if the source of the noise is visible.
Lalal.ai will separate the voice and the noise into two streams so I find by mixing some noise back in at a lower level it's much better that just the isolated voice. There are some settings that can be adjusted but artifacts do show up on the voice alone.
ElevenLabs voice isolation is better but doesnt provide the isolated background noise. However, it is good enough to be used on it's own I think. I still prefer a bit of background noise so my prefered option is to use ElevenLabs voice with a bit of background recorded without anyone speaking provided the fact that is not synced doesnt show up.
This is precisley the kind of AI assistance i would like Vegas to provide. Lalal say there is a VST plugin coming soon.
Thoughts?
