Dutch Goldmine - UPDATE may 26

[audiogeek]

Good recordings and quite interesting that none of the couples seem to be Dutch. Couple 1 sounds Danish, couple 2 I guess from eastern Europe, couple 4 Scandinavian; perhaps Swedish.
Correction is appreciated.

[nod32]

All are dutch. Couple two are Dutch arabics. I guess Turkish.

[Jip]

Heerlijk, nice records nod32

[nod32]

Just incredible, brother! Mind sharing what device are you using? She sounds are sooo clear like daaamn

Just an iPhone for this ones.

[Gerber1990]

Love them, good work

[Seafront]

Wow great recordings! Keep up the good work!

[sndprv]

Wow, the Couple 2 chick was fun to listen to! A vocal, continual moaner.

[findu]

Goed bezig. Nice one

[Robbert]

Thanks for the update and for sharing! I am a Dutch man and live in the Netherlands.

[Fcp045]

Quiet young parents cleaned

Young Parents-edit.mp3

[tufkal]

Erg benieuwd welk hotel het is, kan ik ook eens op pad.

[audiogeek]

Erg benieuwd welk hotel het is, kan ik ook eens op pad.

For the non-native Dutch speaker:
"really curious which hotel this is, can I go there too."

[nod32]

Quiet young parents cleaned

Young Parents-edit.mp3

Really good edit. Is this AI thing public available? Really curious on your cleanup setup. I am self trying to get a sexsound ai detector. But I did not get it to work yet.

[Fcp045]

Quiet young parents cleaned

Young Parents-edit.mp3

Really good edit. Is this AI thing public available? Really curious on your cleanup setup. I am self trying to get a sexsound ai detector. But I did not get it to work yet.

The base model is audiosep-dp which is open source. However, the base model itself doesn't really do a good job with sex sounds because there's nothing NSFW in the training data. This edit is actually from my old pipeline which tried to use audiosep to isolate the noise then subtract it, which works somewhat but isn't that amazing.

So my new pipeline builds Lora adapter training on top of audiosep using synthetic mixtures. Basically you take a section of noise from the track you want to enhance and mix it with a known clean signal of sex sounds and talking - - thankfully we have some beautiful clean exemplars of sex sounds from the Members section. You can do this synthetic mixing on the fly within the training loop. The Lora learns to separate the noise from your target track from your exemplar sex sounds.

If you look at my last few uploads they were done using this method. I'll rerun this capture with the training method later today.

[Fcp045]

Young Parents 1-edit2.mp3

Young Parents 2-edit2.mp3