In the GODIVA link mega b sent

Download the pdf and scroll down to see all the examples

There's some really amazing examples

Honestly it's insane to think that ai can do this much

And the mc one is really interesting to me since the bottom one uses pocket edition, on a spongebob map, with a some mod

Unfortunately, the uberduck.ai bot has not gotten verified by Discord yet even though it should have been by now. It has been taking a long time for Discord to verify the bot for some reason.

those are probably just some random youtube videos

it learned from

Rod Serling needs to be added as a voice

His twilight zone narrating

wrong channel

bottom

Unfortunately, the uberduck.ai bot has not gotten verified by Discord yet even though it should have been by now. It has been taking a long time for Discord to verify the bot for some reason.

?

if it said bottom then this happens 👇

Unfortunately, the uberduck.ai bot has not gotten verified by Discord yet even though it should have been by now. It has been taking a long time for Discord to verify the bot for some reason.

whenever you type starts with a b o t the bot responds lol

Unfortunately, the uberduck.ai bot has not gotten verified by Discord yet even though it should have been by now. It has been taking a long time for Discord to verify the bot for some reason.

Unfortunately, the uberduck.ai bot has not gotten verified by Discord yet even though it should have been by now. It has been taking a long time for Discord to verify the bot for some reason.

bot

this is strange

yeah that's dyno for ya :\

I am thinking of doing the voice of tommy vercetti in TTS Format, there are more than 200 or 500 voice lines, almost 900, which has the voice lines of the mentioned character, so when I feel like it, I will

In order not to exhaust myself, I will take out 256 voice lines

so why do the new voices tend to have a stroke whenever you input more than a certain amount of words like, is the model at a point in training where it knows what the character’s voice sounds like but just can’t assemble it into a cohesive string and the end result is that it knows where to start but it can’t keep up a successful synthesis and so the sound eventually devolves into a stream of audial mush that vaguely resembles the character speaking words

Some characters speak in short sentences normally

so they tend to freak out with long sentences

Anyway, I wanted to experiment but I wonder how people do transcriptions in Arpabet... So, do I follow the conversion format that was used in the Tacotron 2 input or should I remove the pointed brackets?

Also yes, this is the dataset for Mei from One Stormy Night I’m doing!

I’m just gonna experiment putting normal texts up front and then Arpabet in the other, kinda like how LJSpeech does two duplicated in each line in the cvs file, but I replaced the second one with Arpabet inputs