Fifteen.ai Thread

Forums » General Discussion » Fifteen.ai Thread Search Posts
Started by Soft Lava
Posted
20 replies
Subscribe
Soft Lava
Marenheit Contributor - This little pony contributed to the Marenheit fundraiser of 2020
Auti.. Artist -

الصداقة حرام
Twilight is my waifu object of admiration. You have any ideas of what I can do next?

Twilight abandons her friends because she found a greater power. In mcdonalds.

elements of harmony won't come near the power of the elements of hamburager:

pickel. sosig. tomato sloice. onion. beef.

And what element twilight represents? The element of buns of course.
ToastedTruffles

Early Adopter
I had no idea fifteen.ai had left. I had popped in for a bit during a couple of the threads, categorized it as yet another of the incredible-yet-niche MLP projects and moved on. Maybe with bookmarking it for later experimentation.
Soft Lava
Marenheit Contributor - This little pony contributed to the Marenheit fundraiser of 2020
Auti.. Artist -

الصداقة حرام
@ToastedTruffles
Oh, my bad. By that I meant that the site was shut down for a while and was back in business after some time. Well, sadly 15.ai is undergoing maitenance again.

It's sad that 15.ai dev can't just make a standalone project, an executable that works without any need to go to his website. But I guess the process behind voice cloning is so hard that you can't simply make a program that will run on your pc independently. The project might use resourses that are very hard to make an executable of. Either way, that's terrible news to someone who just likes the idea of ponies saying anything you want.
ToastedTruffles

Early Adopter
From what I understand, the person behind the project is using some serious supercomputers to develop it. I have no idea if it needs to be hooked up to these supercomputers in order to process every single text to speech request, or only for building up the voice banks. Either way, the person doing that is someone who's charting new ground in making high fidelity speech, and is specialising in making these speech banks using tiny amounts of sample audio compared to other text to speech projects.
Soft Lava
Marenheit Contributor - This little pony contributed to the Marenheit fundraiser of 2020
Auti.. Artist -

الصداقة حرام
@ToastedTruffles
Yes, the amount of work put into this project is ingenious. From all I know, I've tried to build one github repository that used tenserflow and even a nice user interface. But the results were very bad, although the dev argued the program would clone the voice from 5 seconds of raw data.

The neural net algoritm is primitive and does not rely on a huge amount of data. It's training the ai that does the trick. from the get-go I assumed that once all the training is done there would be an option to share an executable so people would just run it on their hardware and expect something nice in return.

But again, Even a couple of words take minutes to be generated, let alone sentences. there is no doubt that 15.ai heavily relies on a great power if it were capable of firing generated lines with a pace of a shotgun.
AA

Hopeful Pioneer
Who wants to take bets on it coming back online? I don't think it'll be ready in any less than 2 days.
Soft Lava
Marenheit Contributor - This little pony contributed to the Marenheit fundraiser of 2020
Auti.. Artist -

الصداقة حرام
Imagine having this voice cloning technology working the other way around. If, say, you record your own voice line and then use it as a reference for ai to work on. This way, ai doesn't need to improvise in relation to immitating the intonation and emotion of the speaker. It simply changes the modulation of the source file to match the desired result.

To be even more opportunistic, with enough RAM you can then use this technology in real time to prank your friends. you start a zoom meeting and send anything that you say to neural network. Then the ai repeats back what you have just said but with a different voice. This way you can literally be someone else. If this thing becomes real, I wish for someone to use cheerilee's voice to teach shit.
Soft Lava
Marenheit Contributor - This little pony contributed to the Marenheit fundraiser of 2020
Auti.. Artist -

الصداقة حرام
15.ai is still down. The guy is such a snob that he won't let you use his product until he is personally satisfied with it. Ego at its finest. Look at me, I am a big guy. I don't care that you like my project. I an closing it down until I think its good.
AA

Hopeful Pioneer
@Soft Lava
He epitomizes the saying, "Perfect is the enemy of good." I wish he would realize that as long as people are having fun it doesn't matter how perfect it is. Also what stops him from working on it while keeping it up?
Background Pony #3E55
@AA
The reason he says he takes it down is so he can take the GPUs that would be running the site and instead use them to train the next version of his model. Also, I would suggest trying out TalkNet. It is surprisingly good and the ability to match a reference audio is pretty cool.
Syntax quick reference: *bold* _italic_ [spoiler]hide text[/spoiler] @[email protected] +underline+ -strike- ^sup^ ~sub~