Dear Community,

I am looking for a software that can imitate voices by means of artificial intelligence using audio files and play them back via text-to-speech. I have been fascinated by Hans Paetsch since my childhood. He had a beautiful, unique voice and read thousands of fairy tales, literature and children's stories. In Germany, he was and still is considered the nation's "fairy tale uncle" and "the man with the golden voice". Over the years I have collected tons of material from him. Unfortunately, he passed away in 2002 and it would be my absolute dream to revive his voice.

I have already done some research, but I have always come across dubious offers. Either the function was explicitly only available for English voices (I was looking for a German one), or no testing options were offered and credit card details were requested during registration. I also came across some tools that run via Phyton, but unfortunately I lack the technical knowledge to make them work.

My question to you is: Is there any software or any way to implement this idea?

If someone could help me, I would be very grateful.

Many greetings from Germany

HansPaetsch

 

A month or so ago I tried Voice.ai which also gives you an option to train a new voice based of a database (vocal extracts). It is very convincing (at high quality rate but introduce latency) but expensive.

So you could feed it many hours of your guy's voice recording and it would build a database around it.

I made a quick research and they say: ''Our software works with any language, although English provides better results and it's the one we offer support to.''

Might be worth a try. That's the most convincing one I tried so far. I don't remember if there was a text to speech option. But I believe human tonal expression would grant you far superior result so you would need to learn to ''express'' those words like your favourite guy does and speak those words to the best of your ability.

it could be a wonderful journey.

 

Hi. Thanks for your reply. I think, i tried it already and was not satisfied. Does it have a text-to-speech function too? I thought it was only a voice changer which works through real time microphone records.

 

ElevenLabs

https://beta.elevenlabs.io/

 

That's what I said. To my knowledge, there's no text to speech which mean that you get to also learn to express yourself like the man to get it as close as the guy would do it. No text to speech will ever give a human expression behaviour of any human being as accurately as imitating him to the best of one's ability. There's too many factors to take into account. So if you add the voice changer behaviour on top of your interpretation, you will get great result. You need to train the AI with your guys voice first then and only then you will get some decent result. It is not a ''few recording'' training. You need to feed it a very large database. But that shouldn't be an issue for you as you said you have a very large database of his voice recording.

 

This one has realtime function -> GitHub - voicepaw/so-vits-svc-fork: so-vits-svc fork with realtime support, improved interface and more features.

Edit: no txt to speak in this one

 

I can see how AI can reproduce a voice to a degree that's indistinguishable to the real voice but actually reading something, like a story or a poem, requires much more; it has to emote sentences with pace and inflection pertinent to the subject matter. Is AI up to that ability yet?

 

It's getting reasonably close, you can still easily tell it's AI generated, but we're at a point where it'll throw off many people, especially when listening conditions aren't very good.

 

RVC is the best I have used. I used the vocal from a friends song and had Adele sing it. simply amazing.

here are some pre made models. https://huggingface.co/QuickWick/Music-AI-Voices?ssp=1&setlang=en-US&safesearch=moderate

 

Is there any way to train RVC on this voice:

As I said, I have tons of HQ audio material at my disposal. Is it possible to do this in German and how do you run this programme?

 

Yes...you can train any voice.

 

Can you please give me an advise how to do this. Does it work with a software or is it online?

 

https://www.musicradar.com/how-to/ai-vocal-covers

 

ive been using a beta plugin from this team www.rubin.audio its the best Ive heard and works in the daw!

 

sssssssss

This is how I did it.....

 

A lot of prank shit coming from Russian prankers calling to politicians and presidents of EU and US..... No need to imitate voice by themselves....like it was before (if the truth is not blocked in your country)...Just AI shit and all the truth is on the table.

 

Can you provide us that beta program setup file?

 

yes! We are finishing up some bugs right now.. hopefully by August we will have a public beta of the plugin.. and since I've been on AS for a while I'm gonna send it to yall first!!