Hi! I've been a mix engineer for many years, but I've just entered the AI rabbit hole and I'd like to try an experiment.

I have a stem sung by a girl, very beautiful. I'd like her voice replaced with mine.


Is it possible locally with some plugin or software? Or are there cloud-based solutions?

I've seen people on YouTube who make songs with their own voices and then apply the tones of famous pop stars. Very interesting! I'd like to do something similar, but with my own voice instead of a pop star's.


What solutions or tricks do you know?

 

We call it "singing" in these parts.

 

Synthesizer V ---> https://dreamtonics.com/synthesizerv

 

This indeed. I've been using it for the last few months. It is pretty good even though some might say it is still not as good as real vocals. I agree. It is not. But it is better than nothing. Even though it still sounds kinda synthtetic, I can literally get around 95% of what I plan for my vocal melodies.

It is easy to learn if you are familiar with popular DAWs and piano rolls. But, lets be honest, it is not cheap. If I am not wrong I paid around $250 for the studio version and for one additional vocal (it lets you choose one male or female vocal for free but they didn't really work for my songs so I had to buy an additional vocal and use the free ones as background vocals in some songs)

In addition to all these above, if you wanna get a better look, you can check out my band's youtube channel. So far we have released 4 songs and all the vocals are made with Synthesizer V.

https://www.youtube.com/@MilkenTiers

P.S. this band is not an AI band, all the songs and lyrics are written note by note and word by word by real people. For vocals, I record my own vocals first, and then process it through Synthesizer V and tweak it all to my liking.

 

Last edited: Nov 1, 2025 at 2:16 PM

To replace vocals with another, I use Replay or Vocoflex. Replay is free. Vocoflex is expensive.

 

https://podcastle.ai/voice-cloning

https://www.weights.com/replay

 

https://www.ikmultimedia.com/products/resing/

Not out yet, but soon.

 

For offline cloning and inference I'd recommend RVC.

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md

It requires some learnig, preparation, a decent GPU and a few hours to clone a sigle voice (yours in this case) in good quality, though.
Inference, i.e. conversion from input voice to cloned one takes seconds to minutes, depending on the file length and your hardware. There are 2 pre-compiled versions of RVC (for Nvidia and AMD/Intel GPUs) on huggingface, that you simply unzip and you're ready to go. If you know, what you're doing, that is... ;)

Direct links for the official 2023 versions here:
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases/tag/2.2.231006

I would not recommend using voice cloning (neither your own nor anyone else's) online tools for privacy, security and legal reasons.

A quick RVC tutorial here:

https://henrywithu.com/train-your-o...-voice-clones-rvc-web-ui-comprehensive-guide/



PS.

2 more recent, 2024 RVC pre-compiled Nvidia versions here:
https://huggingface.co/lj1995/VoiceConversionWebUI/blob/main/RVC20240604Nvidia.7z
https://huggingface.co/lj1995/VoiceConversionWebUI/blob/main/RVC20240604Nvidia50x0.7z

 

Last edited: Nov 1, 2025 at 5:56 PM

This one does not support voice cloning, I'm afraid.

 

How so? It supports training your own local model on your own material.

 

My bad, seems you're right. Let's wait for the Witches then, and see how well it works. ;)

 

Last edited: Nov 1, 2025 at 6:36 PM