Realivox Blue Lets You Turn Text Into Vocals


RealiTone has released Realivox Blue – a new vocal synthesizer that can translate words you type into vocals.

Realivox Blue is based on 12,000 samples, carefully edited to allow individual sounds to be composited into words and phrases.

Here a video intro to Realivox Blue:

Here’s a collection of audio demos:

Here’s what RealiTone has to say about the new virtual vocalist:

Six vowels and 23 consonants give you the tools to create your own words and phrases. But we went deeper than just sampling a bunch of vowels and consonants. Way deeper. Consider these elements:

An “s” singing into an “ee” sounds different from an “s” singing into an “oo.” (Try it yourself. Listen to how your own “s” will sound different, depending on the vowel.) So each consonant was recorded separately going in and out of each vowel. There are two and a half octaves of “k” samples (32 samples) that go into “ah.” Another 32 of “k” into “eh.” And so on. For each consonant, that’s six sets for each vowel.

When we recorded the ending consonants (completely separate samples), we noticed that it doesn’t sound quite right if you just slap an end consonant sample onto the end of a vowel sample. A singer’s mouth closes differently, depending on which consonant they’re going to. So we sampled not only 192 ending “t” consonants (6 vowels times 32 samples per vowel), but we also recorded the closing vowel sounds leading into each consonant. For ah. And for eh. And ee. And oh. And oo. And uh.

So when you play “Peace,” you’ll be triggering four samples altogether. First is the “p” sample that specifically with “e.” Then you won’t hear just any “ee,” but you’ll actually get the “ee” that comes after “p.” (Your ear would know if we tried to cheat and use a regular “ee” instead.) Then you’ll hear a few milliseconds of a closing “ee” that would lead into “s.” And then finally, the “s”. A closing “s” that specifically goes with an “ee” vowel..

This is why Realivox Blue is 12,000 samples. No other vocal library that we know of has gone this deep in the quest for realistic vocal performances.

We sampled legato intervals for all six vowels, as well as for “mm.” (We find humming to be a particularly useful articulation.) Not just the six/seven vowels, but also the special transitions betweem say, oo to ee. When you sing from oo, to ee, there will be a “w” in the transition. (Try it yourself and you’ll hear what I mean.) So we have “oowee” legato samples. And “oowah.” And “ooweh.” And “eeyoh.” And on and on.

Then we spent months editing and balancing to get the legato as smooth as we could make it.

Realivox Blue works in Kontakt or Kontakt Player, Mac & PC, VST, AU, RTAS, AAX. It’s available now for US $149, with an introductory price of $99.95.

Check out the audio demos and let us know what you think!

21 thoughts on “Realivox Blue Lets You Turn Text Into Vocals

  1. I’ve studied this topic quite a bit, and was pleasantly surprised by how effective this is. Especially the poly legato thing.

    As they mention, there are more than seven vowel sounds, but they seem to have handled the issue of transitioning between vowels & consonants. And at 12,000 samples, I can see how they had to draw the line somewhere.

  2. I’m re-releasing all my Synthtopia comments in a special Collector’s edition – send $10 to reserve your copy! After that it’s onto ‘Wikipedia – the box set.’

  3. Anti-social bedroom producers around the world breathe a sigh of relief.
    Brace yourselves for a torrent of badly produced tracks with samey not-as-robot-sounding-as-before vocals in them….

      1. That’s not funny – it’s my taxes that will be going to support Enya from now on, now that she’s been replaced by this mechanical chanteuse!

  4. Have been using this sound bank for a couple of days and it’s a really nice addition. I did have some script error problems while running Kontakt in Live 9 but using it in Reaper gave no problems at all. No idea why this is. The word builder works quite well as long as you remember to specify your words in a phonetic way. I did notice some noise/hiss which gets quite audible when you process the samples heavily. But overall, it’s a wonderful sound bank that lends itself for hours of experimentation. And yes: you can make her say the nastiest and weirdest things 😛

  5. This one’s impressive capabilities probably won’t get much of an airing. I wish I could find a better word than “sincerity,” but more of these devices than not sound like one-trick ponies. Roland’s VP keyboard update had a major choir in it under total vocoder-mic control, yet the best I ever heard from it was Ed Diaz’s superior cover of that worn-out old fave, “Amazing Grace.” If you’ve heard anything from Vocaloid that doesn’t sound like Daft anime Punk, please post a link. I tend to like the funny initial uses of a novel thing, but the real test of its musicality is how people use it over time. You have to have good ideas and sweat over them to make it rock. You’d have to apply this very carefully not to come off sounding like a poseur. They’ve done a great job, because the demos are very descriptive, so I’ll wait and see if anyone takes it up and sets a performance standard with it. Frankly, I’d prefer to tweak a pitch-shifter into shape and sing things myself. I suspect that would sound as good and be easier to render.

    1. I don’t care for this sort of thing and while the producers have made heroic efforts I agree that it’s going to be better known for its limitations, eg the Amazing Grace rendition here still sounds like a non-English speaker put through Autotune. But it’ll likely get extensive use for backing vocals, on video game soundtracks where the music serves as sonic wallpaper, and for odd effects.

      I wonder why all these products are built around female vocalists? I’ve actually found it a problem to source good sample banks of male vocalists if I want the effect of a choir or something. Sometimes I multitrack my own voice but I don’t have that big a range and can’t hit the low notes. Pitch-stretching with formant correction can help but the artefacts become a lot more noticeable in the lower registers.

  6. I’ve seen many of these types of instruments come and pass, and I’ve bought a few of them… My issue is that they are rarely updated and harder to use than expected. The tools might be present, but the thought that goes into producing with these things is mind numbing. Most of the guys at my studio who use the vocoloid stuff end up shunning it as an actual vocalist/ backup singer and instead use it for strange sound effects. The limitations can be a bit annoying when everything works except one word or articulation.

    That said, this looks to be a bit better than the vocoloid stuff in some respects.

Leave a Reply

Your email address will not be published. Required fields are marked *