/content --------------------------------------------------------------------------- NameError Traceback (most recent call last) in () 24 tacotron2_pretrained_model = "/content/MLPTTS" 25 ---> 26 if not exists(tacotron2_pretrained_model) or force_download_TT2 and using_local_file == False: 27 gdown.download(d+model_ID, tacotron2_pretrained_model, quiet=False); print("Tacotron2 Model Downloaded") 28 elif using_local_file == True: NameError: name 'exists' is not defined

even than i putted "anyone with the link"

NOOOOOOOOOO! after doing my voice, it's only static! because a fail in the trainer...

is there a way i can further train a model?

if you are training tacotron2 you can resume training by just redoing the steps in the notebook and setting output_dir to the model you want to resume training

thanks!

training my new model

that's flower from bfb I think

Finished making a TTS voice of CABAL from Tiberian Sun. not happy with the result.

This is what he should sound like.

its not an easy voice for the ai to replicate

Its a voice that uses vocal doubling and I really havent seen alot of voices where it could imitate that well

so how do i teach the ai to work much like Speaksonia? i am more used to that TTS system because i am trying to use Bonzi buddy as a character in a video game.

Is that even legal lol

Though, tacotron is based on Neural networks imitating speech, so if you have training data on Bonzi buddies speechs pattern you could train it

But even then, youd rather use his TTS engine than Tacotron

yeah i dont really see the point of making an ai replicating bonzis voice when it already exists

^

finished with painty's training so far

results are....

well it has trouble saying words again "let us check below the sea"

this one painty only said half of the setence

"what is happening with my voice"

WIP

fox announcer?

yup, spesifically the one from animation domination

or anidom for short

no i didn't name it that way, fox just called it that

oh, i cee