Pure data patching circle Brussels @ Variable #07

*//*The Patching Circles are open workshop sessions to learn and share
about the pure-data software and other.

Nicolas d’Alessandro will come to present us his approach to speech
synthesis, working on an analog model of a voice flux filtered by
organs, different from classical analytical sample based aproach :

/Hands on Performative Speech Synthesis in Pure Data/

Purposes of voice synthesis have significantly evolved over the last 50
years, tackling more and more complex challenges. The early first
challenge has been to produce intelligible speech. Intelligible means
that the message (words, sentences) can be understood by others. For
this purpose, voice articulation is defined as rules, trajectories on
production models parameters. Later, we see the challenge of producing
voices with a more natural timbre. Here "natural" means that it has to
be close enough from a real human voice, in order to be confusing and
thus comfortable for the listener.

This challenge has been addressed for the last 15 years by using
prerecorded voice material (databases) and reorganize contents (by
concatenating subdivided units) in order to meet the requested sentence.
The impact of these different voice production strategies on achievable
realtime abilities and controllability of the system is huge. Relying on
a voice model gives access to production parameters, which is nice for
realtime control, as far as a meaningful mapping can be designed (e.g.
some problems with physical models). Using database segments helps in
achieving natural sounds but highlights many controllability problems.
Transversally to these issues, the "how can I patch that in Pure Data ?"
question often pops up. This is exactly what we address in this
tutorial! It is useful for anybody who is interested in producing
synthetic voice in a realtime application, a digital instrument or an
interactive artistic project. We will use some basic understanding of
how voice is produced, not more than the basic source/filter
representation, and more complex aspects of voice synthesis (such as
spectral processing, articulatory models, unit management) will be
re-explained and illustrated with examples. Indeed studied techniques
will be described through simple patch examples.


The workshop will be followed by a small performance and a drink,

session open and free to all, debuting as well as veteran,

subscribe please: pdcirclesATogeem.be

_*mailing list: *_

to get updates and participate in the group discussion, you can join the
mailing list here

@ Variable

RueGallait 80
1030 Schaerbeek