Emvoice One is a new virtual vocalist plugin that lets you synthesize vocal tracks, based on the pitches and lyrics that you enter.
The developers describe the results as ‘realistic vocals’. The sound examples sound similar to autotuned vocal tracks.
Here’s what they have to say about their approach:
“Usually, vocal synthesis requires complex synthesis and modelling algorithms that run on your host computer. The results can be impressive, but this technology has not yet reached a level of realism, and has been stagnating for some time.
Emvoice takes a different approach. We’ve broken sung vocals down to the granular level, recording the elements that make up individual phonemes at multiple pitches. Thousands of samples are reconstructed by a sophisticated cloud-based engine that returns the complete vocal to your system over the internet. What you’re hearing when you listen to Emvoice One isn’t artificial – it’s a real singer’s voice interpreting your own words.
The Emvoice One plugin makes it easy to program notes and tie words to them, and the Emvoice Engine does the hard work behind the scenes to recombine phonemes… but there’s one more layer to how Emvoice works. The dictionary translates English-language words into phonemes to more easily speak to the Engine, and also offers multiple pronunciation options for some words, as well as the ability to add your own. Of course, you can skip the middle man and directly type English phonemes into Emvoice One’s text boxes, should you wish.”
Audio Demos:
https://soundcloud.com/emvoice/sets/songs-feat-lucy
Emvoice One In-Depth Demos:
Pricing and Availability
Emvoice One is available now, for Mac and Windows, for $200 USD. A demo version, with a limited range, is available as a free download.
I’d like to see how this compares to Vocaloid 5. It sounds similar but one would think that Yamaha’s far greater experience creating synthesized vocals would give them an advantage in terms of realism. However, since they focus more on the Japanese voices than the English ones I could see Emvoice producing better English vocals than Vocaloid offerings. A side by side comparison would be great!
This and Vocaloid are interesting as an effect, and I can imagine them being used for temp tracks or as a novelty.
But truly realistic vocals aren’t here yet. Google has been doing interesting work, using machine learning to create more realistic speech. If this same approach could be applied to singing, it might get us closer to ‘realistic’.
While not perfect, you should check out synthV as a much more realistic machine learning approach like you said. Also it has a free unlimited trial!! Here’s two links to show what it sounds like:
https://youtu.be/4gAxopnHrJo
https://youtu.be/X5xABAFyh2M
im actually a vocaloid producer, though an unproductive one of late (mostly bedridden by illness a few years)
sounds like a good vocaloid 3 or so to me, but not near the expressiveness of most v4s and even utau, and especially not synthv’s – i like the interface though but i think obfuscating so much in the magical cloud is going to limit any expressive tuning and nuanced control
sounds like avanna
Thanks for the insight. Hope you’re getting better.
And it might be fun to hear from your experiences with Yamaha’s tech, in languages other than Japanese.
Ios app please!
It must be difficult to do vocal simulations without sounding a bit artificial. There may be ways to make these things sound more natural (at the sampling stage)– but it is hard to say. It would be great if the final sound had the sound of a properly recorded vocal with a great vocal mic (like a C12), and where consonants (s, t, f, etc.) sit right in the sound.
I can see why they would want to render offline. But I wonder how that affects the workflow.
— by “render offline” I should have said– “cloud render”
https://soundcloud.com/loremipsumdolorsitamet/01001100-01101111-01110110-01100101a