Yes! Now we can show the result of our first attempt to enhance
Yamaha Vocaloid using our Performance Modeling technology (PM).
Unfortunately, we were unable to get the ready product right at the
NAMM Show, where we had some discussions with Ed Stratton (Zero-G)
and Hideki Kenmochi (Yamaha),
but a week ago Leon Library was delivered to us by express-mail from
Best Service (Germany).
We had the aim to completely exclude manual
editing of Expression elements in Vocaloid itself, which is laborious
enough, and to try to process the initial MIDI file, using one of
Modeler Styles (Onyx Arranger PMStyle). Potentially, our algorithms
are able to automatically generate any number of functions,
contextually dependent on the initial melody. The existing PM
algorithms generate Expression, Pitch Wheel, Modulation, Panning, and
control Velocity, StartTime and Duration of Notes as well.
Unfortunately, from the standard MIDI file, Vocaloid only imports
Velocity, StartTime, Duration and Pitch Wheel. In the future we are
planning to add to our algorithms the generation of synthesis control
parameters, used in Vocaloid (Resonances, Harmonics, Noise,
Brightness, Clearness and Gender Factor).
For the time being, PMStyle was used for the first experiment,
which has done the following automatically (No manual Vocaloid
expression parameters editing at all!):
- Appointed Velocity, according to phrase division.
- Made the notes sound legato within the phrases and inserted the
pauses between the phrases.
- Generated ALL Pitch Wheel sendings (curve), providing
microdeviations, gliding, bending, pitch ornaments and Vibrato.
To analyze the result - the song Sad Mondays (see its details
below), you can take the following files in mp3 and *.mid format:
ALL (6) MIDI files in
Comments by Alexei Ustinov, CTO
Jasmine Music Technology
- Vocaloid surpasses other attempts of vocal synthesis and this
technology undoubtedly has great prospects.
- Vocaloid has extremely flexible synthesis parameters control
system, allowing to realize a huge number of variants of sounding,
including the ones quite similar to real singing.
- Adding the expressiveness while creating a vocal part is the
necessary stage of achieving the sounding close to the real live
- Manual addition of expressiveness elements, available in
Vocaloid, requires laborious work and doesn't guarantee the user the
achievement of satisfactory quality unless he or she doesn't possess
rich musical and acoustical experience in this sphere.
- The flaws of this Vocaloid version are:
- inability to add expressiveness elements in real time mode.
- rather overextended time of rendering
- GUI functionality is not entirely complete
So, Yamaha offers the unique technology, practically unlimited if
we speak of the possible result, but lacking control effective enough
to achieve such result.
As a matter of fact we come across the typical situation relevant
for most synthesis methods: the method itself allows to achieve
various soundings but there are no effective ways to control the
synthesis and the technology resources are underused. For example,
Yamaha VL Syths series, possesses extremely flexible synthesis
parameters control system, allowing to achieve unique soundings. But
neither keyboards (with traditional wheels or joystick), nor Flute,
Sax, Guitar or Violin MIDI-controllers are able to generate necessary
control parameters even with the real performance. The creation of
MIDI sequences that include all necessary commands (using Piano-Roll
and controller's drawing using mouse) also doesn't qualify as
productive. Concerning the VL Synths control I can speak more surely
since I have studied their possibilities myself. I dare say that the
only way to use resources to their maximum potential is our
Performance Modeling technology, providing intellectual and contextual
transformation of MIDI data and enhancing them with performance
nuances in real time. I have no doubt that the developers of VL Series
will agree with me after having listened to the samples, created, for
instance, for MU100R, or after some experimenting with VL Synth, using
our software. I am willing to demonstrate them the sounding of their
device impossible to achieve by any other way and which they have
never heard. I am also sure that Performance Modeling technology is a
real chance to automate the addition of expressive elements to
Vocaloid, which will make it more user friendly and widen the circle
of its users...
You can imagine that this is the way that other control functions
(Brightness, Clearness, Vibrato, etc.) can be generated. And, of
course, the PM algorithms can be created for the various characters of
singing, including "Soul manner" like Stevie Wonder, Whitney Houston,
when one long note is replaced with a number of short slurred notes of
appropriate music scale.
Demo Song details
Tune author - Ravi Dattatreya, Lyrics - Igor Fomchenkov,
Arrangement - Alexei Ustinov
Come on your time is gone
You've been a way too long
Sad Mondays yet to come, my baby
Nothing will be the same
And you will see again
Sad Mondays yet to come
You said you'd have your day
You'd meet your love again
Warmed by the morning sun, my baby
But you forget one thing
No matter what you think
Sad Mondays may yet come
Well after all you'll live through the Sad Mondays
No matter what you'll live through the Sad Mondays
Whatever you try you will see the Sad Mondays again
And again the world will turn 'round
Whatever you try you will see the Sad Mondays
And this is a plain fact of life