28 Aug
2014
28 Aug
'14
11:05 p.m.
On Wed, 27 Aug 2014 10:15:47 +0200
Thorsten Wilms <self(a)thorstenwilms.com> wrote:
On 26.08.2014 19:39, Devin Venable wrote:
That's a more detailed description of what I thought about. I think it
could still be possible to have layers - like in a sampler - but
another editor view would be required. I think also loop points would
work fine. The problem of course would be representation, and, well,
development.
My dream feature? Click on segment, select
"convert to sample" and
a new midi track appears linked to a plugin sampler, ready to play.
I could have used something like that a few times.
Being an architecture astronaut by hobby, I've been wondering about
the differences and commonalities between (audio-)sequencers and
samplers. In both cases you have an abstraction on top of actual
audio-files and playback start/stop. But playback is parameterized
differently. The sequencer has a timeline position and transport
state, while the sampler takes notes (and potentially a bunch of
control events).
Lets think of a track in a sequencer to actually be a mapping of a
playlist to a playback graph. The playlist can be reduced to a
sequence of events. The playback graph consists of at least a
playback engine and may contain an effect chain.
In typical sequencer use, for pure audio tracks, a playlist will
contain regions that map to audio-files. That part of it can be
reduced to a sequence of audio-sample-values, as main input to the
playback engine. There may be other events/automation, all being
inputs to the playback graph. The general idea is that you always
deal with sequences of events as inputs to a playback graph.
Now all of that is coupled to a single, global playback control,
consisting of transport state and timeline position. What if you
could choose to decouple tracks from global control and make them
take note-events for playback control instead? If the playback engine
can variate playback speed in relation to note-value, you have a
basic sampler. If the playback engine also offers realtime
pitch-shifting/time-stretching and formant control, you have
wonderland.
Note that you would not be limited to control a decoupled track's
position and state by notes. There could also be direct control with
start/stop/speed (including negative) and locate/go-to events. Tracks
that play tracks ... Mmwuhahahaha!