The trick with Clarity, from what I've read is that it doesn't try to recognise noise and remove that, it's actually aural network that is trained to recognise discrete human voices and eliminate anything that isn't. It works pretty damn well too.
Except the time when it didn't, and I made Clarity cry. To its credit, it still managed better than what I could muster in RX.
Waves Clarity Vx
-
simon.a.billington simon.a.billington https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=341278
- KVRAF
- 2604 posts since 12 Nov, 2014
-
- KVRAF
- 1583 posts since 26 Aug, 2019
Finally got around to testing with an honest to goodness lawn mower over dialogue clip.
They released it as Non-Native for M1 even though their back catalog is already M1 native as of v13? WTF Waves!!
In Rosetta under Studio One it pegs my CPU at 90% and is unusable. Increased buffer size to 1024 and that solved it. With a single track and dual channel analysis of a single stereo clip it chews up almost 50% CPU according to Studio One.
Works in Reaper under Rosetta without the same CPU issues, so I also tested there. I heard distinct artifacts and voice distortion above 40%. Probably material-dependent what level those kick in, but there is no free lunch. It seems to works best in moderation alongside other similar tools. The best combination I found for the lawnmower noise was Clarity Vx > ERA 6 Noise Remover Pro > gate ... each of them lightly applied, sort of a poor man's Clarity Vx Pro.
Did end up purchasing. Hope they get the CPU performance improved and port to M1 soon. They released such a hog in such a rush.
They released it as Non-Native for M1 even though their back catalog is already M1 native as of v13? WTF Waves!!
In Rosetta under Studio One it pegs my CPU at 90% and is unusable. Increased buffer size to 1024 and that solved it. With a single track and dual channel analysis of a single stereo clip it chews up almost 50% CPU according to Studio One.
Works in Reaper under Rosetta without the same CPU issues, so I also tested there. I heard distinct artifacts and voice distortion above 40%. Probably material-dependent what level those kick in, but there is no free lunch. It seems to works best in moderation alongside other similar tools. The best combination I found for the lawnmower noise was Clarity Vx > ERA 6 Noise Remover Pro > gate ... each of them lightly applied, sort of a poor man's Clarity Vx Pro.
Did end up purchasing. Hope they get the CPU performance improved and port to M1 soon. They released such a hog in such a rush.
-
simon.a.billington simon.a.billington https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=341278
- KVRAF
- 2604 posts since 12 Nov, 2014
Yeah there will be an M1 version on the way. Its very unusual, but my best guess is they began work on it before the "newer" framework was available, so they were limited to developing on the older version.kidslow wrote: Tue Apr 19, 2022 12:11 am Finally got around to testing with an honest to goodness lawn mower over dialogue clip.
They released it as Non-Native for M1 even though their back catalog is already M1 native as of v13? WTF Waves!!
In Rosetta under Studio One it pegs my CPU at 90% and is unusable. Increased buffer size to 1024 and that solved it. With a single track and dual channel analysis of a single stereo clip it chews up almost 50% CPU according to Studio One.
Works in Reaper under Rosetta without the same CPU issues, so I also tested there. I heard distinct artifacts and voice distortion above 40%. Probably material-dependent what level those kick in, but there is no free lunch. It seems to works best in moderation alongside other similar tools. The best combination I found for the lawnmower noise was Clarity Vx > ERA 6 Noise Remover Pro > gate ... each of them lightly applied, sort of a poor man's Clarity Vx Pro.
Did end up purchasing. Hope they get the CPU performance improved and port to M1 soon. They released such a hog in such a rush.
Although, I am surprised its not out already. Perhaps its a neural network thing?!? I know it takes awhile to train those, and if you have to train it again for the M1 infrastructure?!?