Realistic voice transformation?

[Kinh]

Shit, the problem isn't making a male's voice sound like a female, it's pitching it up so it does. You can apply formant in your range and get kinda feminine but the moment you start raising pitch you squeeze the living hell out of it. If that's your intention then you're wasting your time with 'voice changers' because it's transposition that's gonna f#ck it all up. Dont matter if it's melodyne or little alter boy or the IRCAM oscillators, you really gotta learn to sing as high as f#ckin' possible and sound like Lyndsay Buckingham.

[househoppin09]

Just to be clear, my primary interest isn't so much in the usual "male-to-female" thing, more like "male to another male that doesn't sound like a processed version of the same person" or "female to another female that doesn't sound like a processed version of the same person". I'm interested in age, vocal timbre, and laryngeal anatomy transformations much more than the basic gender swapping. At this point I'm about 99% sure that what I'm looking for is impossible with current technology, but if anyone has any ideas, I'm still interested!

[Konduktor]

I'd love to have the ability to realistically voice act multiple characters and in different genders. I don't know of any utilities which can accurately achieve this, but I would think that it wouldn't be too far-fetched with current technology.

There are already a lot of apps which do text-to-speech synthesis, and vice versa. Normally, the text-to-speech voices sound flat and emotionless. But it wouldn't surprise me if someone created a similar utility, only it would detect the syllables, pitch, etc. in your own voice, then "replace" it with one of the built-in voices. That way, it would be a different voice, but retain the same words and emotion.

[Kobamoto]

any new info on this subject, ircam trax v3, or any other app that is good for changing male vocs into female vocs and vice versa???

[Richard deHove]

Not revolutionary but gets pretty clean results:

[Ah_Dziz]

Toneboosters Voice Pitcher. Then something with a formant shifter like steinberg pitch corrector or one of the many Melda pitch or fft based processors. This can give (in my opinion one) of the best vocal transformations around. It works amazingly well for spoken vocals. For sung vocals you run into trouble and may have to pull off some more trickery to get back to the original key, or change the key of the song. If you are into max/msp or any of the other in depth modular environments you can do lots of great stuff with wavelets transforms also, but it's trickier to set up and still may give pretty synthetic sounding results.

Another great trick (but expensive, labor intensive, and requiring multiple people) is to have multiple takes of the exact same material done by people with completely different voices, and then use Zynaptiq Morph to mix between them. This only works if the performances are perfectly lined up so you'll be doing editing to get everything lined up perfectly (either manually or using melodyne or something) so that all the transients, plosives, and siblance are in the same spots. Once this is set up though you can get a ton of completely distinct voices saying or sing in the same thing and it can sound extremely natural while not sounding like any of the source voices.

Good luck
JJ

[danielh3410]

I don't think there's anything new (or all that effective). For years I've periodically looked fairly thoroughly for something like this, but have always come up empty. I don't think that vocal-alternation technology has advanced sufficiently far yet.

That being said, if I've missed something, I'd be happy to be proven wrong!

I watched a grantouch piano demo about 20 years ago and the voice changing tech was amazing-a man sang into the machine that was the electronic brains inside the grantouch electronic piano and it changed his voice into tina turner

[danielh3410]

I watched a grantouch piano demo about 20 years ago and the voice changing tech was amazing-a man sang into the machine that was the electronic brains inside the grantouch electronic piano and it changed his voice into tina turner

[Belinski]

Flux Trax is very good for voice manipulation within +/- 12 semi tons, you can transform gender, age, and play with the articulation of the voice (amplify or reduce), without too much artefact. You can easily transform a single voice or a group of voice like a choir.
Izotope Vocalsynth2 is not too bad but the vocoder doesn't sound that good.

[dandezebra]

Dehumaniser can do some pretty neat stuff. Not sure it is what would help as it tends to transform sounds into... The nonmelodic?

[cbmtrx]

Based on my experience, the best vocal transformations are done *manually* using a waveform editing app and then a bespoke plugin chain that achieves the sound you're looking for—or one that completely surprises you.

I've never found any single plugin that can do enough by itself, so I experiment with many. It's generally trial-and-error, very time-consuming, and sometimes you'll come up empty, but every so often you'll have created something wonderfully unique.

[Obsolete462444]

A cool way to do it:

1.) pitch-shift the backing instrumental you are singing on down a few keys (using the included offline pitch-shifting algos in your DAW) and import this backing instrumental into a separate project

2.) record your vocals over the pitched and slowed down instrumental (obviously you need to sing in tune of the pitched down notes)

3. Now take those recorded vocals, bring it back into the original instrumental arrangement and "reverse-pitch" (for example: if the backing instrumental was pitched down initially by -3 semitones, now that you are back in the original instrumental arrangement, you need to pitch the vocals you recorded back up by +3 semitones).

You can try this using different quality pitch-shifting algorithms (some act like tape, where pitch and time are related variables, some change pitch without affecting the timing constants) - experiment and try different combinations.