# Recode

# 1 Concept

Alice and Bob have a conversation, where Alice’s live voice is replaced by Bob’s recorded voice, and Bob’s live voice is replaced by Alice’s recorded voice. The replacement is at best imperfect, but hopefully Alice and Bob can communicate on some level.

# 2 Setup

  • computer running Debian with JACK

  • full duplex stereo sound card configured XRUN-free at low latency

  • two hand-held vocal microphones, connected to left and right input channels

  • left and right output channels connected to stereo speaker system

  • the recoda-live program running during the performance

  • two spoken word performers

# 3 Score

Performer A uses the left microphone.

Performer B uses the right microphone.

They improvise a dialog, each leaving pauses for the other to respond.

They should pay attention to the sound from the speakers more than any direct sound they are able to hear.

Themes to include mimicry and repetition with variation.

Towards the end of the performance, both should say the phrase

I think we may have been using the wrong microphones.

Then finally they should swap microphones, so that A uses right and B uses left, and then repeat the phrase in unison.

# 4 Performances

  • Even Salon +- either/or 2016-04-30 was performed by Jonathan Kemp and Claude Heiland-Allen.

  • Recode FM 2019-12-17 was performed by two analogue radios and streamed live online.

  • Election night mashup 2024-07-04 22:00 UTC+1 was performed by two digital radio internet streams and streamed live online.

# 5 Implementation

code.mathr.co.uk/recode

git clone https://code.mathr.co.uk/recode.git

Computes energy per octave (using FFT) of each live input chunk, and outputs the best matching chunk recorded from the other channel. Recently used chunks are penalized to avoid buzz.