These are AI voice things are indistinguishable from the real person and while this is funny as hell it just means that absolutely no recordings are dependable anymore.
Along with other issues, the most notable is that the voices are incredibly flat, and set to a singular emotion.
Minute irregularities in audio that can be detected when looking at the data stream itself, in conjunction with a failure to properly apply emotion to speech is how they will be caught. The latter would literally require an A.I. to study an entire persons life in order to determine how they would react to certain information, and apply the correct emotions. Which means the only time this will be a problem is when people are assisting such programs to assure that the proper emotions are applied, and data irregularities are cleaned up.
Remember a good chunk of the supposed "Nazi warcrimes caught on picture" have turned out to be fakes made through image splicing with an artist covering up the obvious splices. We've always been a position where we can't trust things at first glance.
I wonder if you could use a FFT on the voice sample and see a difference between a computer generated voice and a computer captured voice. It would take someone more familiar with voice capturing systems, data compression methods, and voice reproduction systems than me.
Possibly, mind you I'm only barely aware of how FFT functions. That being said, if all your training data comes from say a news anchor from the news show specifically, the ML model will train not just on the voice, but on the background noise as well. Which means the noise accompanying the voice can be a giveaway if say the questionable audio in question was supposedly not from a news broadcast, but has the same background noise that accompanies real news clips.
Anyone got a link to one with a non-robotic voice, that can handle large amounts of text and allows audio download? E-book in...audio out is what I'm looking for.
[ + ] Crackinjokes
[ - ] Crackinjokes 5 points 1.2 yearsJan 31, 2023 13:51:11 ago (+5/-0)
[ + ] s23erdctfvyg
[ - ] s23erdctfvyg 0 points 1.2 yearsJan 31, 2023 15:38:46 ago (+0/-0)
Minute irregularities in audio that can be detected when looking at the data stream itself, in conjunction with a failure to properly apply emotion to speech is how they will be caught. The latter would literally require an A.I. to study an entire persons life in order to determine how they would react to certain information, and apply the correct emotions.
Which means the only time this will be a problem is when people are assisting such programs to assure that the proper emotions are applied, and data irregularities are cleaned up.
Remember a good chunk of the supposed "Nazi warcrimes caught on picture" have turned out to be fakes made through image splicing with an artist covering up the obvious splices. We've always been a position where we can't trust things at first glance.
[ + ] PotatoWhisperer2
[ - ] PotatoWhisperer2 0 points 1.2 yearsJan 31, 2023 22:52:56 ago (+0/-0)
[ + ] s23erdctfvyg
[ - ] s23erdctfvyg 0 points 1.2 yearsJan 31, 2023 23:36:18 ago (+0/-0)
[ + ] drhitler
[ - ] drhitler 0 points 1.2 yearsJan 31, 2023 19:00:59 ago (+0/-0)
[ + ] Hoobeejoo
[ - ] Hoobeejoo 3 points 1.2 yearsJan 31, 2023 09:28:21 ago (+3/-0)
[ + ]deleted
[ - ] deleted 2 points 1.2 yearsJan 31, 2023 08:28:07 ago (+2/-0)
[ + ] Clubberlang
[ - ] Clubberlang 1 point 1.2 yearsJan 31, 2023 09:39:40 ago (+1/-0)
I always likes the way kagaroofuckers pronounced garage
Gare / ridge
[ + ] deleted
[ - ] deleted 1 point 1.2 yearsJan 31, 2023 09:55:45 ago (+1/-0)
[ + ] Shotinthedark
[ - ] Shotinthedark 1 point 1.2 yearsJan 31, 2023 19:03:02 ago (+1/-0)
[ + ] Clubberlang
[ - ] Clubberlang 0 points 1.2 yearsJan 31, 2023 10:43:09 ago (+0/-0)
[ + ] dontbeaphaggot
[ - ] dontbeaphaggot 1 point 1.2 yearsJan 31, 2023 11:53:22 ago (+1/-0)
[ + ] ModernGuilt
[ - ] ModernGuilt 1 point 1.2 yearsJan 31, 2023 09:05:14 ago (+1/-0)
[ + ] TheGreatWhiteHope
[ - ] TheGreatWhiteHope 3 points 1.2 yearsJan 31, 2023 13:23:24 ago (+3/-0)
https://archive.4plebs.org/pol/thread/414310894
[ + ] beanbagWizard
[ - ] beanbagWizard [op] 0 points 1.2 yearsJan 31, 2023 18:00:22 ago (+0/-0)
[ + ] Jiggggg
[ - ] Jiggggg 1 point 1.2 yearsJan 31, 2023 08:14:48 ago (+1/-0)
[ + ] ImplicationOverReason
[ - ] ImplicationOverReason 0 points 1.2 yearsJan 31, 2023 12:59:38 ago (+0/-0)
Anyone got a link to one with a non-robotic voice, that can handle large amounts of text and allows audio download? E-book in...audio out is what I'm looking for.
[ + ] MichaelStewart
[ - ] MichaelStewart 0 points 1.2 yearsJan 31, 2023 20:21:14 ago (+0/-0)
[ + ] deleted
[ - ] deleted 0 points 1.2 yearsJan 31, 2023 23:39:58 ago (+0/-0)
[ + ] Clubberlang
[ - ] Clubberlang 0 points 1.2 yearsJan 31, 2023 09:38:23 ago (+0/-0)