Did you have to type the words in with phonetics in mind or is it that smart? Like if you wanted to say "though" do you type in "tho" so it doesn't want to say "thou"?
Think about rough and through. Rough may be interpreted as ruff or as rouf, but if the program treats "ough" as "uff" like in rough all the time, through would become thruff. Dough, or "duff"? Is it that intuitive to know the difference? If it gets those words right that would lead me to believe there is a dictionary in there somewhere.
Think about read. It can be pronounced "reed" or "red". Lead is the same way. "Leed" or "led". You can LEAD off first base, pronounced LEED, or you can solder with LEAD solder, pronounced led. Interested in knowing how the interpretation engine handles words like that.
"The Browns LEAD 21-10." Something we don't hear much around here...
Nope, no phoneme notation needed for this little project. It has them available, but I didn't have to use any. Nor did I have to do any disambiguation of words.
There is definitely a dictionary. If you type "glorkparm" it puts a red line through it and will not proceed until you resolve it to a known word or phoneme. It is also, to whatever extent it can, using complete recorded words – this is sort of a recording-based system of (gag) "RealVocals".
As for the learning curve, I'm sure there's more to learn here, but not a whole lot. Here's all it took to create this line in Logic:
– 1) Insert Emvoice plugin on track with MIDI melody to be sung – 2) Play piece in real time so plugin can record melody – 3) Type sentences into blobs
Apart from that, I had to do a little blob-sliding to adjust phrasing, to determine which notes are separate and which are part of a single sung phrase (staccato / legato). But that's all. It was REALLY easy.
Demo version lets you play with a limited pitch range.
I was able to make out about 75% of the lyrics. The really wordy phrases I couldn't get. Not bad software though. Not worth the money for me personally but very good performance.
That's a hoot... amazing technology. I can't imagine ever using this on a project, but I'd bet it would be fun to play around with. Thanks for posting Mark.
It's a pleasant sound, but - as other people have noted - intelligibility can be low.
The Emvoices seem to work well when the syllable rate is fairly slow.
They also have a glassy, over-processed tone, so they work much better with a production that uses that sort of tone.
With the progress of vocal synthesis and neural networks, we're rapidly approaching the point where synthetic voices is virtually turnkey. I expect to see a lot more high-quality synthetic singing programs in the near future.
In fairness to Emvoice, I didn't try to "help" at all, this is just what you get when you type stuff in! I find the least intelligible parts are repeated notes in the extreme high registers, and where odd complex phrases would be hard to decipher anyway (like "hot pepper bear spray tears streaming face covered all in sweat" or "why did he go don't know he quiet", which is talking-to-a-dog talk.)
For the record, here be the words (I deliberately omitted them to see if they were unnecessary, sorry!)
I went to turn off my television set When suddenly he caught my eye Hot pepper bear spray tears streaming Face covered all in sweat The insurrection had my guy
Why did he go Don't know He quiet Left me alone to stare and cry After the Capitol riot I sold him to the FBI
When I first heard this I was a bit dubious about it, but I was tempted by it. Having just watched the Youtube videos (Emvoice) on how to use it they can have my money right now. Finally I can have a singer for my songs. Its not going to put any vocalists out of work but its certainly going to work for me.
Edit: Having purchased the full versions I am more than happy with the product. I did have one query so I sent an email to support. This was on a Sunday morning and I had a reply in less than 5 minutes, which clarified my question.
In fairness to Emvoice, I didn't try to "help" at all, this is just what you get when you type stuff in! I find the least intelligible parts are repeated notes in the extreme high registers, and where odd complex phrases would be hard to decipher anyway (like "hot pepper bear spray tears streaming face covered all in sweat" or "why did he go don't know he quiet", which is talking-to-a-dog talk.)
Yeah, the problem with higher registers is that it's harder to make out the formant, so there's a lot more room for confusion about the phonemes. This is the sort of problem that sopranos run into all the time.
Add to that a low-level of predictability for the lyrics, and comprehension rapidly disappears, no matter how articulate the singer.
But I was also comparing some of the officials Emvoice demos when I was commenting on the voice quality of Emvoice. The voices are pleasant, but very synthetic. They remind me of the old Vocalwriter program - which I really liked the sound of.
Here's an example of someone using Emvoice to create choral music:
It sounds quite good, but the clarity isn't there.
Then again, choirs aren't ordinarily understandable, are they?
I have no doubt that Emvoice will only get better as time goes on. In the meantime, just don't print your lyrics to keep the haters in the dark.
I have heard some truly amazing productions posted here over the years using such technology. By “amazing” I’m trying to articulate how well the melody was presented, the manner of phrasing and how easy it was to understand the lyric. My caveat is that I’ve yet to hear one for which it is not immediately obvious that the vocal is synthetic. The best I’ve heard let that work for them. Once soul and other emotive aspects become apparent it will indeed be a new day.
My caveat is that I’ve yet to hear one for which it is not immediately obvious that the vocal is synthetic. The best I’ve heard let that work for them.
That's my thinking. Yes, you can tell it's synthetic, so let's not try for convincing imitation or we'll just land in the "uncanny valley". I'm hoping to play more with things only a synth can sing.
My caveat is that I’ve yet to hear one for which it is not immediately obvious that the vocal is synthetic. The best I’ve heard let that work for them.
That's my thinking. Yes, you can tell it's synthetic, so let's not try for convincing imitation or we'll just land in the "uncanny valley". I'm hoping to play more with things only a synth can sing.
LOL...so now you will need the copyright on your voice.
I don't think we are very far away from very realistic-sounding synthetic vocals. Five years? Perhaps less. The speed with which technology is advancing is mind-boggling.
I have flown on this airplane at supersonic speeds.
Noise and cost is an issue that is currently being solved. We may in fact have another one in the, not too distant future.
There have also been some serious "over-promising" and "under-delivering". IBM Watson has had a very hard time in the medical field.
My guess is that looking out across a major city twenty years from now may be a very different picture.
Things tend to be evolutionary and not revolutionary but twenty years ago many things that exist today would have been hard to imagine.
We live in a very strange world/time. People on one street are setting under a bridge smoking crack while a few blocks away people are driving robotic cars around on Mars.
Many wonderful things will come to pass. IF...we don't destroy ourselves.
One of the reasons I think we will have great sounding software-generated vocals is money. There is a huge demand for such a product. Let's call it Voice In A Box. You get Whitney Houston, Sam Cook, and Taylor Swift for $1500. I would buy that in a hot heartbeat...lol
Hell, I would even pay good money for Brent Mason...lol
Of course, I want an animated vocalist to go along with the deal.
Pure science fiction of my younger days is now so common no one even looks up to see it. If you look around very much you will realize the future is already here.
I started this thread and have been using Emvoice's first two voices happily.
This announcement worries me, it seems like it could herald a nasty step backwards:
Quote:
Meet our third voice, Thomas. Inspired by the ever-expanding world of electronic dance music, Thomas captures the unmistakable classic vocoder sound that defined a generation. The only difference is that it’s way more convenient to use: no equipment, no vocal performance skills, no mixing wizardry necessary.
Of all the things I could have asked for in a new voice, "classic vocoder sound that defined a generation" would not have been one of them.
Fascinating.