My prediction was correct, you admitted defeat and you failed to provide a better sounding "Real-Time Harmony Vocal Effect VST audio example" that was recorded with VST's in Real-Time with your voice. (no DAW offline created harmonies or hardware harmony vocal processor created harmonies)

it's very amusing how you conveniently avoided my challenge with an excuse and instead you took the easy route, and you just stated the beyond obvious suggestion in your message "simply record more tracks with someone actually singing the harmonies"

Thanks for enlightening me and everyone else in this forum thread with your extensive knowledge and insight of creating harmonies, because nobody else would have ever thought of trying your brilliant and complex idea of "simply record more tracks with someone actually singing the harmonies" (Sarcasm)

 

Last edited: Jan 31, 2024

whatever dude...

 

THIS DOG SHIT SANDWICH TASTES BETTER THAN ANY OTHER SHIT SANDWICH AND YOU CAN'T PROVE IT DOESN'T

Why not make a sandwich out of something other than shit

YOU ARE DEFEATED HAAAA

 

Lately I've been writing a lot of demo songs to sell, and sometimes I don't even have time to record my own harmonies.
I record a double track and build harmonies from it using melodyne. I leave the first one untouched in time, but the subsequent ones I detune in time and a little in pitch (Add random deviations, time drastic) having previously selected all the notes in the editor. Then I export everything and replace the voices via "replay". So, it seems to me that it sounds more varied and there is no distortion after manipulating the notes.

 

Could be useful to have some sort of randomize button or pot for this type of harmony creation purpose- you know, to adjust the length, pitch etc a little bit. Or is there already somewhere?

 

You can basically run AI stuff like that on big servers or locally on your own computer. While I'm not super experienced, I figured that locally seems more save in terms of other people getting hold on your own personal voice models (if you want to make one) as well as well as that websites can change , also possibly since regulations about what is allowed when it comes to vocal AI could change.

I used this youtube tutorial mentioned below to guide me through installing the softare needed and it explains how to use it. With this specific software you can either train your own voice model (own voice, celeb voice or whatever) and you can also import voice models others have already made. These can be found on the internet too (for free).

Here you can download the Mangio-RVC Ai voice software to make the voice model or use existing voice model.
https://github.com/Mangio621/Mangio-RVC-Fork/releases

This video guides you through how to install it and how to use it

On the following sites you can download voice models (That you may mix with eachother too, to get a new voice)
https://docs.google.com/spreadsheet...F51ZSynOvEjz8fHqwriX1wUMPw/edit#gid=537336172
https://www.weights.gg/nl
https://huggingface.co/models
https://voice-models.com/

If you want to blend voice models with eachother, it can be done with this software

https://huggingface.co/wok000/vcclient000/tree/main

This video explains how to intall and how to use this to blend different RVC voice models with eachother, basically to create a new voice. I think this can be used if you do not want a celeb in your song, but do want some voice. The more voices you blend, the less it will sound like anyone in particular.

If you any other questions I might be able to asnwer them or might not be, since I'm pretty new to all of this Ai voice stuff as well. For example, I managed to install the first software where you can either use already trained RVC1 or RVC2 models from internet as well as train them yourself using any sound source, but did not train one myself. I also tested the voice blending which worked well. I blendeddfor example freddy mercury models with James hetfield, and the new indeed sounds like it's right between these 2 voices.

As I said above, I first used the Dreamtonics Syntheszier V VST to create singing with their voice Solaria. I believe there is a cracked older version without VST option, but I just bought the software (the first software I bought in a long time, well worth it!) because it feels groundbreaking to finally have easy acces to a good voice that basically is as easy as creating a melod in the midi editor (pianoroll). Just now you also write words on top of it, that become the lyrics. If you plan to buy it, I would advice anyone to buy the voice 'solaria' rather than any other voice, as it's the most realistic and the best singere of the bunch (most Synth V users said the same)

Dreamtonics, the company behind Synthesizer V keep updating their softare Synthesizer for free free, and each time interesting new features are included. The most recent update allowed was Voice to MIDI, which allowed you to use any wav and convert it into MIDI that solaria then sings, without having to do this manually, meaning you can just feed your own song with your voice to Synth V and it automatically creates all the notes and lyrics into the Synth V voice, so now Solaria can sing it. I personally do not use this feature, but it's cool they added it.

So for anyone interested in getting a convenient voice in their music

1. Create melody in DAW piano roll and import that into synth V or use their Voice to Midi option (only newest version)
2. Then if you created the melody manually, write the words of lyrics in synth V software
3. If you want another voice than the synth V voice, then export the voice wav you just made in your DAW into the first AI software I mentioned (Mangio-RVC)
4. Then use that AI voice software (Mangio-RVC) to alter Solaria's voice into any voice you want. You can pick a model online, train one yourself on your own computer or blend different ones with the other Ai voice software
5. Then if you want to change that voice, you can use Ai voice blending software to basically blend one voice with the other, which basically creates a new voice. You can do this as often as you want, although I do not know if this lowers the quality (eventually)

Disclaimer: While I've tried all of the above myself , the Ai stuff changes rapidly and there might be more recent AI software that might do a better job. I tried all of this like 2 months ago. so keep that in mind. This guy on youtube (Jarods Journey) does good explanations, so if you want to be kept updated about voice AI stuff, his channel is pretty good for that.

While I signed up 12 years ago for this website I'm a noob (Kinda absend minded though, as I forgot I even had an account here recently or even forgot that I knew this website 12 years ago), so it was just by pure luck I was reading this specific thread again which I made me see someone (you) had asked me something. Maybe you (or someone else) can tell me on this website how you know someone replied to you, like you did to me? Or are you not notified at all when someone makes a reply ? Thanks!

 

Last edited: Jan 31, 2024

With Antares and others you will sound very close to a chipmunk from them adding an effect to your voice and not allowing a pre-delay adjustment per voice. I tried all the harmonizer programs. Most old school harmonizers don't allow manual pre-delay adjustments per voice and also uses processed voice effects and synthesized voices. Any voice processed by any effect will affect the original sound and have you sounding close to the chipmunks from personal experience. I tried them all and the realest harmonizers just multiply your actual voice without adding anything to it. For hardware TC Helicon had a rackmount Harmonizer called the TC Helicon VoiceDoubler which multiplied your voice four times over which was really good. The main thing to sound realistic is that you also need predelay separation for each voice to sound realistic. Most others aside from Nectar 4, and Eventide Quadravox don't allow your own predelay editing, as they're always set at predetermined predelays ensuring you remain in that chipmunk realm. A true harmonizer will allow you to just multiply your real uneffected vocals and add ability to adjust the pre-delay to make it sound like you in four different voices. Only a few actually does that.

 

Thanks 0on3 firmware received. I already have version 1.009 Build 26 installed - as you can see. When I run the custom update in Soundsupport2 and choose 'vlive_dsp_f130.syx' it asks for the build number of the firmware (do you know the number???). If I just hit enter, it says the firmware data is not valid.

Also, you supplied another file called 'Update_VoiceLive_f130.exe' but that file loads an attached .htm help screen implying it's for the Pedal version - see bottom image.

Update: I found the source of the files you sent they are from here under 'Voicelive' https://www.tc-helicon.com/downloads.html#googtrans(en|en)

The rack version does not list any firmware at all on their site. SoundSupport2 lists the version I already have. Still, I find it strange that you can get the VL Pedal editor to work for the rack. I'll leave it at that and hope for a response to this. Thanks for all your help 0on3.

 

Last edited: Jan 31, 2024

Those features and preferences are all under Alerts. https://audiosex.pro/account/alert-preferences.

 

Semi - automatically, there is a button, but it is in the drop-down list menu.
With the notes selected, go to edit -> Add random deviations -> timing: Drastic

Or use Revoice Pro to generate a double from a modified voice with melodyne to add difference, and then use this sample to replace the voice. But for me it's faster through Melodyne.

 

Can you share an example here?

 

What's this all about?
Trivox Harmonizer- TriVox features three harmony voices that can be precisely tuned anywhere from -12 semitones to +12 semitones.
https://www.waproduction.com/plugin...term=6522168403091&utm_campaign=6482821326491

 

Trivox
Here's a lil tut for tat:

 

I'll try to send it a little later

 

Last edited by a moderator: Feb 3, 2024

The only thing I have found that truly works convincingly (this is the keyword here) is tweaking thru melodyne. There is something to be said for a vocal effect being introduced, but that is an effect and will sound like an effect and not like real vocal tracks. Taking a base track and reproducing it x number of times for each harmony and then running melodyne on each subsequent harmony track - pitching, slightly de-tuning, and slightly de-timing each one fore or aft will produce REAL sounding results without actually recording each track with the talent. It can all be done in the box this way and give truly convincing results, if that's what you're after. With melodyne you have complete control over what harmonies you want to create. However, with more complex vocal/harmony lines it might just be faster to have the talent record the tracks and tweak the minutiae later. Having good vocal talent can make all the difference in this respect.

 

I have not trained voice models myself, as I did not have anyone specific in mind to train and I am not a singer myself. I was just experimenting with the software and used other parts of it (using created singers are blending them) rather than training them. However, the video that I posted (first one) was about the same AI doing voice cloning. So you can do that with that, but I just have not tried that option myself yet, casue I had no need for it (yet).