I'm just gonna let it run while I watch some netflix or something

dont

oh, how soon?

in 2 minutes ..?

I am very new to machine learning in general, so thanks for the help

and colab isn't letting me use a GPU, wonderful stuff

if I had a good computer I'd just do this on my own PC, but I don't

just wait

it'll work again

This is beautiful.

has anyone tried to train other languages by modifying the symbols file and got decent results? just want to know if i can get on that

alright, this isn't exactly what I wanted, help? did I not train it for long enough?

i am very new to this stuff

what happens if you train it for a year

it'll probably reach a peak and stay there until it's finished or, it will reach a peak and then slowly go down in quality until the AI can only say the phrases you inputted

not enough training data probably

you waste a lot a lot of money and resources 🤷 and the resulting voice will be much worse than if you trained just for 30 minutes

it won't stay at it's peak

yes

83 files, over 3 minutes of data... Is that not enough? There exists no more data.

it's not a lot

but the voice model is also not that bad

Hmmh, I trained the voice a couple more times, if it's better I'll submit one of those. If not, I'll just submit the first one

Is this your voice?

it's the stellaris xenophobic advisor voice

Ok, so I had a question. Say I wanted to use a character like Shaggy. How would I express emotion through the bot's voice ?

It needs a separate dataset for each emotion.

But that takes longer and will require to modify the model and code a bit more, like 15.ai's when it attempts to apply with DeepEmoji for contextual synthesis.

Think of the rapper datasets for example: 2Pac has speaking, calm and intense. That's three datasets.