Really looking for help with an Austrian Painter Everyday / Calm Voice model, included is the raw audio to train it. Primarily sourced from the secret Mannerheim recording of a private conversation with Hitler in 1942, which is not particularly good quality but absolutely priceless nonetheless. There are only two other examples of how he spoke around normal people and situations, which is of course totally different than his boisterous oratory vocal style we've come to associate him most strongly with. Included in audio is every example of him speaking casually I can find, making this should be fairly straightforward for anyone familiar with the process of ai model creation
#Austrian Painter Everyday / Calm Voice
shy nova
fathom anchorBOT
hardy bobcat
https://1drv.ms/u/c/038509ea3e2b01a2/Ec75iZJiM9JPhvtoCDdKjIEBhorGlPc8_tX9GPexNpm...
The microphone is not consitent enough, if trained, it would be less likely sound like him
the early 10 secs clip sounds really clear for something that was recorded during that time, but from there onwards it became kind of reverby which probably going to make the model sounded he's talking from a can
shy nova
the early 10 secs clip sounds really clear for something that was recorded durin...
My previous experiment with this audio to train turned out rather well
but I no longer have this model, unfortunately
shy nova
where it gone?
The external harddrive I had it on went where the red fern grows, went up in blue magic smoke and died
shy nova
also did you get the dataset from YT?
I can't recall, my archive is sourced from everywhere
hardy bobcat
try removing the noise and train it in .wav audio format
I could try training it but like it's hard to differentiate if the voice is speaking was hitler or other person because the calm voice is not that recognizeable
shy nova
it's all Hitler, I spent a long time with the transcript cutting all Mannerheim's audio
shy nova
try removing the noise and train it in .wav audio format
I don't even have any software to do this, my previous model was done with Elevenlabs voice cloning
hardy bobcat
it's all Hitler, I spent a long time with the transcript cutting all Mannerheim'...
What did you use to cut the audio?
shy nova
Audacity, I tried pulling the background train noises but it ended up turning out really reedy and tinny every time I tried it, I went ahead with the raw cut clips and the quality of the product was good enough that I left it as-is
hardy bobcat
do you perhaps have or saved it as .wav after being cut
shy nova
that MP3 is how I exported it, the file can be converted but that MP3 is the original
hardy bobcat
Oh
so I'm gonna assume it was a one time edit and you didn't save it as Audacity project file
shy nova
Nope