Dreamtonics announced their new project “Vocoflex” today.
Vocoflex is a real-time voice morphing plugin which can transform a vocal recording to sound like one or more other voices, including the ability to combine voices.
As shown in the demonstration, this can be combined with Synthesizer V Studio to create vocal sequences as if you were the one singing them, rather than Kevin, Solaria, etc. Bear in mind this is making the SynthV output sound like your voice, not the other way around (that’s what Voice-to-MIDI is for).
To avoid unethical uses of the technology, the output will include an inaudible watermark which can be used to trace a sample back to the user who created it. This mitigates the risk of copyright violations, nonconsensual use of someone’s voice or property, and impersonation. Based on the diagram shown in the video, the watermark should be robust enough to still be detected after the mixing process, without compromising the quality of the output.
The plugin is currently in beta. Unlike with SynthV Studio, this is not an open beta, and includes a signup and screening process before access is granted. You can apply for the beta here: Vocoflex Beta Testing Program Signup Form (typeform.com)
Full information about Vocoflex, including an FAQs section, can be found on its new page on the Dreamtonics website: Vocoflex | Dreamtonics株式会社
(please note that I’m still not associated with Dreamtonics in any way, just sharing the news!)
Thanks for this awesome news!
It seems to be unique with multiple voices.
Exactly what is expected to improve voices (and AI voices).
The watermark seems to be a good thing to fingerprint.
But I am wondering how dreamtonics can prove that the origin of the sample file is compromised afterwards. They saved source files on their own servers to prove it?
The FAQ says that all rendering is done on the user’s device in an offline manner. The watermark likely includes a unique ID that ties back to the user’s license, so that Dreamtonics can trace it back to the person who created it if necessary.
Dreamtonics can trace it back to the person who created it if necessary.
I understand that but it is not enough to know the user. You need to know if the original source file is really coming from the corresponding voice. It is the only way to prove it: The source file!
I doubt false accusations are a realistic thing to be concerned about here. Anyone who is working with someone else’s voice will have a contract or written permission proving that there was an agreement made.
If you can’t prove you have the rights to use a sample, you wouldn’t be able to use it in a song anyway.
This is a very theoretical way of looking at things.
Especially when we see the multiple AI voice sites that allow learning to take place (some already have 34!).
It goes without saying that at some point some voices will sound more and more alike. So it is the original of the voice which will be the proof. Especially since if we combine voices, all this will become more and more complex to prove, if not impossible.
And the trials will multiply! Only in case of success of course.
Yeah, it’s certainly not a perfect solution. The watermark doesn’t help for many situations, like impersonation over the phone where most people can’t or don’t record the audio.
I think the idea is that there will always be a tool capable of these things. Dreamtonics can’t stop it from happening entirely, but they can take steps to make scammers and harmful individuals use something other than Vocoflex by adding barriers to unethical use.
I am glad to see they’ve included some safety measures from the start, but I’m also not entirely convinced they’ve taken enough precautions.
I agree with you on this point. They do something on this particular security point and it is the minimum. May be adding a check box (source file is legal?) before uploading a file to involve the user a little more.
also recognizing that nefarious actors will use this and other similar tech however they want regardless. worse for some folks will be if their license is stolen and used by bad actors. i think it’s nice for law abiding folks but like any tech, it will be used for warfare by state and non-state organizations looking to do bad things.
that said, it looks really cool and i’m hoping i can get on the beta (and ultimately get a copy once it’s live.)
I know a number of songwriters who have started using SynthesizerV because they are older and their voices are no longer in prime condition. Since they’ve got access to their recordings, it would be easy for them to create legal vocals that sound more like them than off the shelf SynthesizerV voices.
However, how well that works depends on how closely aligned the base vocal is to their own vocal.
In the long run, I’d like to see this embedded into SynthesizerV, because I prefer to work on the vocal in the standalone version instead of in the DAW.
They’ve only just announced it, and I’m already complaining!
Hello. Very excited with this new program Vocoflex. Currently I have Synthesizer V and 3 bank voices. Do Vocoflex will work with my current 3 bank voices or also will include the Lite bank voices or any others?. Thank you.
So do we need any other voice banks for SynthV? Doesn’t Dreamtonics shoot themselves in the foot?
Maybe they saw voicechangers will be here anyway, so why don’t do their own and lead the way.
This solves my last problem with SynthV - using voices I like.
I can’t say with certainty, because I haven’t used it. So take anything I say with a large grain of salt.
But RVC leaves the underlying performance such as the pitch line and the duration of the phonemes in place.
So if the source vocal doesn’t have the same mannerisms as the target singer, it won’t sound much like a vocal clone of the voice. At least, that’s been what I’ve heard in various examples of RVC vocal cloning.
