 |
Wim, a lot of people won't
be familiar with you and your products. Please give us a little bit of your background in brief words.
|
 | I come from an academic
background and have been working as an audio researcher for eight years before starting the company. It all started
at the Visionlab of Antwerp University where prof. Dirk van Dyck gave me the opportunity to do a research project on
musical signal processing. To do things properly, I applied for a postgraduate course in acoustics, signal processing
and computer science applied on music at IRCAM in Paris (IRCAM: Institut de Recherche et Coordination, Acoustique et
Musique / Research and Coordination Institute on Acoustics and Music).
I was lucky to be one of the 16 students who were accepted out of 150 candidates. During this one year course I did a
project in IRCAM's Analysis/Synthesis team with Xavier Rodet and found funding to start a PhD. After a total of 2
years and a half at IRCAM, I returned to Antwerp and finished my PhD. Then I continued working there for another
three years as a postdoc before I started up Mu Technologies in August 2007.
|
|
Mu Voice is your latest piece of work. Please describe it in your own words.
|
|
It is my first piece of work as well ;-). Mu Voice for me, is like the product I always wanted to have but was
never able to find. As a brass player (trombone) I always wanted to have a real-time piece of software or plug-in
that would allow me to create multiple notes from the one I am playing and to do effects on them. I've never
understood why guitar players use tons of effects while for many other instruments and vocals this is still kind
of taboo. At Mu we believe that Mu Voice is a product that can inspire musicians and singers to start doing this.
Secondly, Mu Voice can be a real money saver for your studio work. If you look at how much time you need to make
the harmony arrangement, practice the parts and record them, you can save many studio hours by using Mu Voice.
You only need the chords and a decent vocal track and the plug-in takes it from there. You can even decide to add
some harmonies in the postproduction stage long after all singers have gone home or do harmonies in a remix.
|
|
What's the main technique behind Mu Voice?
|
|
Mu Voice is based on a new spectral analysis technique that I developed at the end of my research years.
The patent on this technique is very close to being granted so we can prove and protect its novelty. Analysing
the spectrum of a large time window is quite easy because sinusoidal components are not overlapping. The main
drawback however is that the quality is poor and the latency is high. We have found an efficient way to take
into account the overlap between spectral components which allows us to use smaller windows and hence achieve
a higher quality and lower latency.
|
|
Looking at software like Vielklang, Avox and not to forget Melodyne,
there are already some very powerful applications to process vocals out there. How is Mu Voice different?
|
|
Yes we noticed those ;-) In fact, some of these products came out just before we did our first release.
Fortunately, Mu Voice is significantly different on quite a few levels.
A first key difference is that there is a different notion for those products of the term "real-time".
Some of the products you mention can be considered samplers in which you first record the sound, then
perform some sort of analysis off-line, and then, during playback, you can alter the characteristics of
the sound in real-time. So the synthesis part is often real-time, the analysis part however is not. Therefore,
these are not the kind of products a performer could use on stage.
In Mu Voice, both the analysis and synthesis are real-time and on top of that the latency is about 5.6 ms.
These constraints are very challenging to meet. The analysis, synthesis and effect processing must be carried
out on a very short time frame with very little data. In addition you don't want to consume to much CPU. Therefore
the complete system was optimised with SIMD instructions (a special instruction set on Intel processors and PowerPC's
for vector manipulations) for optimal performance.
A second differentiation I think is the innovative product concept itself. Harmonising is a relatively complex task.
How many voices do you want? What effect do you want to apply on which voice? What is the current key and chord? How
can you change these over time? The chord scheme is kind of the key component if you are talking about user friendliness.
Just put the chords and the presets in the scheme, sequence it to the host and there you go: instant harmonies. No
tedious and time consuming pitch curve editing.
Finally, we believe strongly in further integration of new functionality as opposed to developing a bundle of plug-ins
with limited functionality. This gives you better performance and latency if you want to combine different things. Our
aim is to create a single solution that is as complete as possible.
|
|
The key feature of Mu Voice seems to be to take into account the musical context
of a song. Please tell us some more about that.
|
|
Mu Voice has a unique way to define harmonies. Harmonies are configured as:
- second voice: about a third above the lead
- third voice: about a fifth above the lead
- fourth voice: the octave above the lead or modifiers like 6, 7 or 9
For each combination of scale and chord we have created a lookup table that is used to
determine which note the second and third voice should sing for a given note in the lead.
Taking into account this information is what we mean by "the musical context".
|
|
Is Mu Voice primarily a software for the studio or can the live
musician benefit from it?
|
|
We aim at being successful both in the studio and the live domain. For the live domain I think we simply have not
seen another native plug-in yet that offers the combination of quality, performance and low latency that we offer.
If you want to process live vocals with Ableton, MainStage or Max/MSP, Mu Voice would be the obvious choice.
As mentioned before the bullet point for studio use is its user friendliness. No more tedious pitch curve editing.
Just configuring chords and presets.
|
|
Especially on stage the singer needs fast results and doesn't want
to toy around with a computer too much. What does Mu Voice offer require to prepare for the gig?
|
|
Well, I guess the Ableton guys must have heard that a lot ;-) Given the success they have had over the past
years, I do think there is a large community already of musicians that use computers for live performance.
For somebody who has a recording setup already, the step of taking Mu Voice to the stage is very small.
The setup that we use for live performance consists of
- A computer (Mac Book Pro, 2GHz)
- An audio interface with both audio and MIDI (MOTU Ultralite Firewire)
- Active studio monitors (Genelec 8020a)
- A MIDI Controller (Roland RD-500, but much simpler MIDI keyboard is just fine)
Especially the choice of your audio interface is important. We have tried some interfaces that were just not
able to perform at the low latency that you would like to use Mu Voice live (and they were not cheap ones).
So "try before you buy" would be the best advice I can give. The choice for Genelec was not only determined
by the sound quality but mainly by the protection that they have over the speakers. This is something you
wouldn't care about for a studio but if you travel a lot with them you want to make sure they can handle it.
(No, we are not endorsed by any of these companies.)
|
|
Processing voice right on stage has been problematic in the past,
because people don't see what's going on. Electronic music seems to change this preconception these days.
What are your thoughts about all this?
|
|
I don't think that people necessarily need to understand what's going on. When I look at a guitar player I
do not really care what effects he is using, it just has to sound great.
One thing I hear from time to time is that people think that vocal processing software can make lousy vocals
sound great. This however is a misconception. Lousy vocals will produce lousy harmonies. My colleague refers
to this as "garbage in, garbage out". The great thing about Mu Voice is that it can multiply a great performer.
|
|
Mu Voice seems to be your only product at the moment. What more can we expect
in the future? Do you plan to specialize on the field of vocal processing?
|
|
Yes, we will remain active in the vocal processing domain. We will continue to extend Mu Voice with
new features and new functionality. One thing that is coming soon is extended MIDI support. Currently
we only process note on and note off events. In the next version you will be able to assign a MIDI
controller to a parameter and be able to navigate through the chord scheme with program change events.
This will allow you to use a MIDI foot controller which is important for a sax player for instance who
cannot use his hands to control the plug-in.
Secondly, some of Mu Voice's functionality will be elaborated further. Especially for the filtering and
harmonic EQ-ing, the current interface is limited to the raw basics where you can just select from a given
set of effects. In the next versions you will be able to create your own filters and make them change over
time. Other things that can be extended are spectral warping curves, pitch warping curves, etc.
As far as other products go, there is a lot of new development in the pipeline as well. We are adapting our
core technology to offline processing and synthesis technology. In this stage we are still working on the
proof of concepts. It probably is too soon to talk about product development. The good news is that Gilles
Dandelooy has joined the company since August and he is currently working on this. He was already helping
out with a lot of stuff. The demo song for instance is his composition. I'm glad to have him on board.
|
