Is there any tool or research out there to encode audio such that a human listener can understand comfortably, but a speech-to-text system cant understand?
Anti-speech recognition
Other urls found in this thread:
arxiv.org
twitter.com
welsh
fucking lost
This
Most languages besides English.
Grow up an unironic cajun like me.
This would be useful because then every youtuber could just run their video through this “filter” and then the audio is (ideally) maybe a little off but then it cant be easily sentiment analyzed by google.
Bump
AAAAAHAHAHAHA
My sides
This goes to Jow Forums humour
Put me in the cap, user!
Hewwwooooo Jow Forums! ^.^
No, you have to use end-to-end encryption.
The point of the op isnt to prevent someone who isnt directly trying to listen to your videos specifically. Just to frustrate mass surveillance
Unironically a very important invention if it ever happens
"adversarial noise" is a common tactic used in reverse deep-learning. It's how recaptcha hasn't been solved by bots already. whether or not the same thing exists for audio, i don't know but you'd likely have to implement it yourself in python
nevermind, here you go:
arxiv.org
tl;dr: the tiniest amount of perturbation applied to the waveform will confuse a speech-to-text algorithm. it's almost like you didn't google this before posting.....
>in python
Aus! Aus!
Hermione used engorgio on her tits
Thanks. I did but had a hard time finding anything, was getting all kinds of other shit. Google has gotten worse for sure.
Will never happen. Whatever sound you can understand, a algorithm only trained for this purpose can understand it too.
>how do we frustrate mass surveillance?
Simple: just nationalize all the filthy corporations doing it, user.