[LAU] The future of audio plugins ?

[Tito Latini]

> For deep learning, the analysis tool itself would be the DAW in this
> case which has context: All tracks, random data access, and it can
> control the plugins.

I think a robot that controls a DAW is an alternative and simplest
solution to avoid a complex DAW with specific AI, shared memory, etc
for a network of plugins. For example, the robot knows how to read the
session file: it reads that file to discover the network of the
used plugins.

The robot knows how to control the DAW (i.e. MIDI and/or OSC messages
for transport control, [un]bypass the plugins, move faders, panpot,
etc), therefore it changes some parameters and gets feedback with
audio data (i.e. from master output). It analyzes the output and sends
messages to mute tracks, bypass plugins, alter parameters and
automation). Play, stop, rewind, continue and jump. The CPU limits the
speed.

In practice, it is necessary at least:

    1) A comprehensible session file
       (simplification: external conversion to a standard session file).

    2) A controllable DAW (MIDI, OSC, etc).

    3) Algorithms for artificial listening: culture, aesthetics and
       psychoacoustic are important elements and the complexity depends
       on the requirements.

The "artistic" sensibility is difficult to simulate (also for humans),
however I think that robot is probably a friendly (*) assistant for a
person without experience.

A simple case: the human requires a mix without to use equalizers or
dynamics processing. The robot knows the style of the music a priori
or it is able to discover the musical genre. If the robot works fine,
the human get a decent automatic mix and, possibly, some suggestions
about applicable EQ, dynamics processing, re-record some tracks, the
alternative mic to use, etc. The human asks to apply some DSP, the
robot provides a new mix, etc..


(*) "ha un cuore di bambino che non si rompe ma a a i"

       https://www.youtube.com/watch?v=zDzieioM05g