this is the first graph i've seen that actually looks like a spectrogram

> Will be extended to a multi-speaker TTS. 👀

uberduck mobile is getting realer by the second

or well by the paper

wow iphone

honestly though imagine spongebob uhhh reading your ebooks or whatever it is that people use built-in tts for still

or better yet, if gpt-j-stuff distillation somehow ends up being extremely effective people could potentially run ai·dungeon-likes entirely locally combine that with special finetuning to put the character's name before each line of dialogue and a massive omnispeaker model with all the sliders one could ever ask for and i would definitely know what i'd take to a deserted island

also the model is slighter bigger than the paper

24M vs 21.8M in the paper

although the pretrained model is not uploaded yet on google drive

eye phon

ngl iphones are hella expensive for what they offer

gamer

can somebody help me train my Markiplier dataset?

i'm getting a "No such file or directory: 'filelists/clipper_train_filelist.txt" when running "Check Data" step

did you run the step where you put in the model name and all

yes

if you restarted you need to run it again

k

i'll try that

@User now after restarting it says the filelist/list.txt files does not exists, but it does

I have a question... What does "Warning! Reached max decoder steps" means? Sometimes I receive this message when I test my model (the AI voice reads the sentence very well, but makes weird sounds at the end), sometimes I don't. Does someone know about what the problem could be?

How hard is it to learn how to train nodels

And learn how to work with GANS

Please @ me

it means the model didn't press the "okay i'm done" button, instead continuing to generate audio until it reached the hard limit

make sure your input has a punctuation mark at the end, or two or more for safety

same goes for the training data