Yoo! Thank you @WillTheWeirdo I had no
Idea of this one! Preciate you!

 

re: Hush.AI was hoping to test it, but its mac only

 

Hello,

Can you add your own custom voices in there? Like if you have something like spoken dialogue from a podcast would that work?

 

c'est fou. thank curtified.

 

add my voice it's possible?

 

Credit where credit is due.
The OP always posts songs that are technically well constructed and produced.
Far better than I could ever do.

What makes them hard for me to listen to personally (any why I don't usually comment) are the usual vocals and vocal style that accompany his songs, and whilst the OP admits to not having the greatest voice, he has sometimes struggled to hear feedback that agrees with that view.

It's interesting therefore to hear new versions of his songs without that technical distraction.

The vocals certainly inherit the tone of the AI artist, but they don't bring any of the original vocal style of the artist to the party (think vocal gymnastics).

So for these songs, they are passable, and non-offensive but ultimately sound emotionless due to the quality of the original source vocal.

Where this technology will really come into it's own is where the OP can simply input the lyrics provided by his mate, and the AI can provide the synthesised vocal sang in the style of the original artist. But at that point, your DAW will probably have a button called
"generate new track and post to soundcloud" and where's the skill and artistry in that?

 

This is probably not the best way to get people to download and install software that you have developed!
That's enough to scare some people

 

Isn't this basically what Yamaha Vocaloid, some of the East/West plugins, and also Dreamtonics Synthesizer V do? If somebody merged text to speech + midi file x Vocoder or other Vocal Synth/Autotune type options, that would be great. So much time and programming efforts getting poured into making AI voices sound more "Realistic"; when the first thing to process them would be to hide the fact they even came from some AI in the first place. Cut out the realism middle-man. If we need Frank Sinatra, we can just go dig him up.

 

The first voice emulation i ever used was this and it still has great charm:

Who said voice emulation has to be realistic?

 

No worries.. Its not like were charging to use the free tool we built..

 

I am going to make a model of this for Replay

 

Wavosaur produced a free plugin emulating that voice, but I’d love to use a Replay model

https://blog.wavosaur.com/text-to-speech-vst-vst-speek/

 

I have shows this weekend.. but ill make a model with this mid next week

 

Completely offtopic: You can have that. Drummers at your command and nothing but at your command. Get Jamstix. It's deep, but you ultimately you will get what you wished for. Yes, sometimes you can get what you want!

 

Yes you can, you need to train a model with your voice (or any voice, as long as you got a clean source to train). I'm not sure if you can train a model with Replay directly, but with something like RVC Web UI (it's an offline tool, so you need a decent GPU to use it), there's some other, but I use this one with great result.

Replay is really great and simple to use (and very quick), I didn't realise you where behind it @curtified, great job, thanks for sharing it for free

I already find pretty usefull way to use it, making "impossible cover/collab" is fun, but it can also be use in a more professional way. For exemple, I train a model using an artist voice I work with, using good recording from old studio session. Now he can send me a voice recorded from his phone, and with just a little cleaning I can make it sound like it's from a studio, make new background vocal, make more natural harmony (using melodyne or replay pitch change) and with replay it will sound natural, not like a pitched voice.

It's one use among many others, it still a new tool, but as you said, it's the worst this tech will ever be.

Let's remember how Midjourney, Dall-E, Stable Diffusion and other image generation were 18 month ago, now we have video/3d generation that are way better than what those not moving image were. And it's just the begining, same with music generation.

It's really exiting to see what will come, and mostly how creative people are going to use those tool, because one thing is sure, AI will not replace the need for creative people to create something new (but it will probably replace/change the way music is consume, probably starting with the sample pack business, the same way image generator didn't eliminate the need for people to paint/take photo, but more and more replace the need for stock image).

 

Sounds amazing!

 

I wish I could type more words into it. Im having a hard time. If we could get a large set of .wav files it would make training easier. Is anyone down to help out with making a pack of sounds generated out of this..

 

You can use replicate for one free model. then after that its like 1-2 bux..

https://replicate.com/replicate/train-rvc-model

Also thanks for the kind words! and shared sentiment!

My goal is there will be a Technologist role in the music making process. Just like a songwriter, mixing, mastering engineer.

There will be someone there to help translate a song into multiple languages, make AI models, etc etc.

 

Marie and Per Back together one more time Thanks to https://www.tryreplay.io/

 

Roxette-Marie-Fredriksson-Necessary (Long Version)

Roxette-Marie-Fredriksson-Me And You And Everything In Between

Marie and Per Back together one more time Thanks to https://www.tryreplay.io/