-
m-relay
<c0mmando:hackliberty.org> like
paste.hackliberty.org
-
m-relay
<monerobull:matrix.org> very good video on haveno
youtube.com/watch?v=kHnPukvGT7E
-
m-relay
<korgprivacy:matrix.org> 📰Missed Monerotopia Episode (#171)? Check out the Price, NEWS, GUEST Segment.
-
m-relay
<korgprivacy:matrix.org> Reports here! ⤵️
-
m-relay
<korgprivacy:matrix.org> 📊Price Report:
-
m-relay
<korgprivacy:matrix.org> Youtube:
youtu.be/4vDoYxxv7ow
-
m-relay
-
m-relay
<korgprivacy:matrix.org> 📰News Segment:
-
m-relay
<korgprivacy:matrix.org> Youtube:
youtu.be/A2xfYLVXwsw
-
m-relay
-
m-relay
<korgprivacy:matrix.org> 👥GUEST Segment:
-
m-relay
<korgprivacy:matrix.org> Youtube:
youtu.be/GJdbwdsLDe8
-
m-relay
-
m-relay
<pepe86frog:matrix.org> I really like your services homie
-
m-relay
<rottenwheel:kernal.eu> Anime? No thanks. Distorted, spooky, odd voice? No thanks. Hard pass.
-
m-relay
<monerobull:monero.social> you can also clone the voice of manly men if you are more into that
-
m-relay
<mmxxx:monero.social> don't put ideas in his head...
-
m-relay
<mcneb10:envs.net> yeah good luck finding software that can anonymize your voice without making sound like garbage
-
m-relay
<mcneb10:envs.net> unless you can figure out how to use so-VITS-SVC
-
m-relay
<mcneb10:envs.net> im going to try that
-
m-relay
<monerobull:matrix.org> literally the software i suggested
-
m-relay
<monerobull:matrix.org> although i dont know if you can maybe reverse it by training a model for it
-
m-relay
<mcneb10:envs.net> thats why you train the model yourself
-
m-relay
<mcneb10:envs.net> im sure it can be reversed
-
m-relay
<monerobull:matrix.org> as in, take the cloned voice & the gura model, somehow get the original audio back out
-
m-relay
<mcneb10:envs.net> unless you are the only person with the model
-
m-relay
<mcneb10:envs.net> even then it probably can be
-
m-relay
<mcneb10:envs.net> probably it would be best to use so-vits-svc + an extra layer to add distortion to the audio
-
m-relay
<monerobull:matrix.org> yeah if you have the original model and nobody else has, i dont think it would be possible unless the process of training models themselves is reproducible
-
m-relay
<mcneb10:envs.net> yeah thats a bit of a stretch but we don't know what's out there
-
m-relay
<mcneb10:envs.net> training a model is also extremely time and energy consuming
-
m-relay
<mcneb10:envs.net> it took 1 day per epoch for me on my K80
-
m-relay
<mcneb10:envs.net> they told me you need ~2000 for the best quality
-
m-relay
-
m-relay
<mcneb10:envs.net> its probably just easier to do voice -> whisper -> tts in real time but then it sounds awful
-
m-relay
<udeeyyyvvv:matrix.org> DEMO AND REVIEWS
matrix.to/#/%23newwwwwcpp%3Amatrix.org
-
m-relay
<recanman:kernal.eu> plowsof:
-
m-relay
<monerobull:matrix.org> better to rent gpu by the hour
-
m-relay
<monerobull:matrix.org> spend a few dollars, probably cheaper than training on your own electricity
-
m-relay
<mcneb10:envs.net> definetly but its cool to self host it
-
m-relay
<mcneb10:envs.net> when i did it i modified a powersuply and spliced the cables to plug it in to the GPU so it would have its own power
-
m-relay
<mcneb10:envs.net> and it still overloaded the PSU