back to
時の崖_tokinogake
We’ve updated our Terms of Use to reflect our new entity name and address. You can review the changes here.
We’ve updated our Terms of Use. You can review the changes here.

AToGRNN Prism

by RM Francis

/
  • Streaming + Download

    Includes unlimited streaming via the free Bandcamp app, plus high-quality download in MP3, FLAC and more.
    Purchasable with gift card

      ¥500 JPY  or more

     

1.
2.
3.

about

AToGRNN Prism re-presents live sets and studio compositions from 2019-2020, recordings of which were fed as input to a recurrent neural network (RNN). The resultant material— which is not heard directly, but serves as input for the subsequent processing stage — exhibits two predominant states: a turgid chugging sound, which is the more prevalent of the two, and silence. Within and between these two baseline modes, a chaotic liminal space is constituted intermittently by transient phrases that briefly flail into existence before collapsing back into one of the two default patterns.


A chorus of 46 unique MIR-based machine-listening agents was constructed to reinterpret and perform the RNN’s output. Each agent was designed to listen to one or more perceptual features of the sounds produced by the network, such as loudness, spectral entropy, etc.; the weight and application of each feature’s data varied from agent to agent. These virtual perceptual-expressive systems were then endowed with a vocabulary of 413 waveforms. Each agent reproduced the RNN files according to its individual perceptual biases using nearest-neighbor matching of the waveforms with frames of the RNN material. Much of the material produced by the network that sounded redundant or static to human hearing contained information of perceptual relevance to these virtual auditors: seemingly indistinguishable passages yielded translations of radically different timbres; stretches of eventless inertia were sometimes converted into rippling cadenzas.


In AToGRNN Prism the predominant compositional movement is not of motifs flowing into one another, but emerging from and subsiding back into a substrate of noise, eschewing macrotemporal development altogether. Refractive simultaneity is emphasized over syntactic coherence. Events materialize singly or in clusters of near-clones distributed across the stereo field as multiple agents apply their own sensory templates and elastic temporality to the unheard source file.

SampleRNN by Soroush Mehri et al.
SCMIR library by Nick Collins
nuPG by Marcin Pietruszewski
FluCoMa library by Owen Green, Gerard Roma, Pierre Alexandre Tremblay
Mastered by Jack Callahan

credits

released November 25, 2022

Visual
Takashi Aoki / takashiaoki.com

license

all rights reserved

tags

about

RM Francis Seattle, Washington

contact / help

Contact RM Francis

Streaming and
Download help

Shipping and returns

Redeem code

Report this album or account

If you like RM Francis, you may also like: