[ / / / / / / / / / / / / / ] [ dir / random / abcu / ebon / k / komica / miku / nofap / random / ytc ]

/hypno/ - Hypnochan

Obedience Will Bring Pleasure
Email
Comment *
File
Password (Randomized for file and post deletion; you may also set your own.)
Archive
* = required field[▶ Show post options & limits]
Confused? See the FAQ.
Embed
(replaces files and can be used instead)
Oekaki
Show oekaki applet
(replaces files and can be used instead)
Options

Allowed file types:jpg, jpeg, gif, png, webm, mp4, swf, pdf
Max filesize is 16 MB.
Max image dimensions are 15000 x 15000.
You may upload 5 per post.


We have entered a new age of filters on /hypno/.

File: 5c8f447c4b3bb2e⋯.jpg (183.25 KB, 1852x1425, 1852:1425, DMmAPqbWkAAE69G.jpg)

File: 1a17bf9d9fd6153⋯.mp4 (966.58 KB, 1280x720, 16:9, ponies_are_uncancelled.mp4)

bfa060  No.64003

Alright, it's time to crash this plane with no survivors.

You can now take an arbitrary 5 or more seconds of someone talking, plug it into a free program, and provided your GPU is good enough I doubt mine is use that person's voice as TTS without training a new neural network model or any stupid shit like that. That's right, you can have a custom hypnosis script voiced by anyone you want, for free.

Here is the GitHub project:

**https://**github.com/CorentinJ/Real-Time-Voice-Cloning

Here are some audio samples:

**https://**google.github.io/tacotron/publications/speaker_adaptation/

And here is a web program with several preset voices you can use if you are lazy or satisfied with a couple meme voices like GLaDOS or ponies (vid related is a sample from it, you can get better results by fucking with spelling and punctuation):

**https://**fifteen.ai/app

Enjoy!

____________________________
Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

bfa060  No.64004

YouTube embed. Click thumbnail to play.

And here is a video demonstration to prove that, yes, it really is that easy.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

b9d44f  No.64007

This kinda makes we want to train a NN to as many hypno scripts as i can find, put whatever illogical mess it spits out through this and give it a try

Cant be that bad ;D

Fuck, i might just do it, already found some textgen projects on github

What a time to be alive

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

9748c6  No.64010

Super happy that ai voice cloning is popular at the moment so we are finally going to have tts software that surpasses over 10 year old ivona tts voices.

At the moment though the voices sound a little bit too distorted and unnatural compared to the best ivona voices, unless you spend a lot of time tweaking and cleaning up short bits of the audio, which would take too much time when you are doing a 30+ minute hypnosis audio. And it's hard to format these ai voices in a long hypnosis script, while it is relatively easy to format using something like balabolka and sapi5 voices. These ai voices are pretty much only made to create very short voice clips at the moment.

But I'm looking forward to seeing someone put together a good tts writing interface once the voices are good enough.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

9f7388  No.64011

>>64010

it has to be better to have a fucked up 1$463||4 V4|ent1n3 voice than the microsoft sam shit tts is now

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

9748c6  No.64013

>>64011

microsoft sam is definitely dogshit, but voices like ivona's joanna or salli are pretty natural-sounding.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

4da8d0  No.64015

Thanks for posting about this, op. I have been very interested in these ai generated tts recently.

Do you know how good the GPU has to be? Mine is modern but fairly low end.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

4da8d0  No.64016

>>64010

>unless you spend a lot of time tweaking and cleaning up short bits of the audio, which would take too much time when you are doing a 30+ minute hypnosis audio

Sounds like an actually really good excuse for people to charge for their files.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

000000  No.64019

>>64003

Do you have any instructions on how to install this on Windows?

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

bfa060  No.64020

>>64015

>>64019

After installing Python, it comes with a commandline package manager called pip. You use this to install pytorch through the instructions at https://pytorch.org/get-started/locally/ , then you download the github repo and follow the readme's quick start instructions.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

c646a9  No.64023

>>64007

>This kinda makes we want to train a NN to as many hypno scripts as i can find, put whatever illogical mess it spits out through this and give it a try

>Cant be that bad ;D

Ha! Was going to do that as well! But, there are not enough scripts available to train a text model. I would say you need at least 100MB. I tried with 35MB - while it can produce something, it's within a couple of phrases. Then the sense is lost.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

d2f358  No.64062

>>64023

Well, damn

Guess i'll go looking for scripts then, maybe start a thread asking for them and sharing the results in return….?

Way too much work hours thanks to COVID19 though, could be a while

Thanks for the info! ;D

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

c646a9  No.64070

>>64062

But if you want something real, you have to look at 100+GB, like the latest chatbot Meena

https://ai.googleblog.com/2020/01/towards-conversational-agent-that-can.html

"The Meena model has 2.6 billion parameters and is trained on 341 GB of text, filtered from public domain social media conversations. Compared to an existing state-of-the-art generative model, OpenAI GPT-2, Meena has 1.7x greater model capacity and was trained on 8.5x more data."

GPT-2 is worth looking at. It's quite good and was trained on "only" 40GB of text.

Have a look here: https://talktotransformer.com/

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

29d0fd  No.64086

>>64003

I had an idea for a program like this. I don't have the will, or expertise to create it though.

The idea of having a library of scripts and voice packs is appealing though. Especially considering you could for instance insert triggers, create infinite loops, add specific details, create own scripts altogether or bring back retired authors.

I think it would have use even for students and audiobook fans for instance.

The problem is the voice intonation though.

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

2bfdec  No.64211

apparently K31 made some neural network files last year

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

98e987  No.64845

anon on /pnd/ found another resource

https://www.voicery.com/

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

af90aa  No.64847

can someone help me install this thing? I keep geeting endless errors and now its complaining about not having a cudNN dll and i'm confused…

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.

69715d  No.64849

for those with a Mac >> https://www.techjunkie.com/make-your-mac-talk-say-command/

The voices in 10.14 and up are outstanding. And you can load lots of languages.

Think different :-)

Disclaimer: this post and the subject matter and contents thereof - text, media, or otherwise - do not necessarily reflect the views of the 8kun administration.



[Return][Go to top][Catalog][Nerve Center][Random][Post a Reply]
Delete Post [ ]
[]
[ / / / / / / / / / / / / / ] [ dir / random / abcu / ebon / k / komica / miku / nofap / random / ytc ]