5.1 Solresol dictionary
We used the Solresol dictionary defined by François Sudre in 1866 and translated by Garrison Osteen in 2011 (see: here). We edited this to remove different word options (i.e. to provide only one meaning per Solresol word), as well as to remove any punctuation and capitalisation. This lists 2660 words, but there are actually: 7 + (7x7) + (7x7x7) + (7x7x7x7) = 2800 possible ’notewords’, i.e. 140 that Sudre did not define; if no match is found, a nearby word is chosen.
5.2 Software
The software is created using Jonathan Wilkes’ ’Purr Data‘, version 2.9.0. This is based on Pd-l2ork (created by Ico Bukvic), which is based on Pd-extended (by Hans-Christoph Steiner). Pd-extended builds on the core Pure Data (version 0.48.0), copyrighted by Miller Puckette and others. Purr data incorporates a number of other packages that have been used, notably ’GEM‘ – the Graphics Environment for Multimedia. The pitch analysis is done with the ’sigmund.pd’ object from Miller Puckette.
5.3 Bird song files
Blackbird: https://freesound.org/people/squashy555/sounds/341675/ - CC0 Public Domain
Skylark:https://freesound.org/people/Benboncan/sounds/72752/- CC BY 3.0
Song Thrush:https://freesound.org/people/inchadney/sounds/117103/- CC BY 3.0
Otherwise, our own recordings were used (see method below).
5.4 Midi Player / sound font
The Pure Data Midi output is sent to the CoolSoft freeware virtualMIDISynth by Claudio Nicora. (This uses the Farfisa piano sound font by Riccardo Loi, the FluidR3 GM bank sound font and the GeneralUser sound font from S Christian Collins (see here for links to these).
5.5 Text font
The Enochian font is freeware created by the Nu Isis Working Group. An alternative font from Wfonts.com was also used.
5.6 Birdsong recordings for offline playback
The device works best with high-quality recordings that isolate the birdsong and have minimal background noise. This necessitates a quiet environment, sensitive microphones, and a high-quality recorder. FEL Communications Ltd (Sandown, UK) ’Clippy’ omnidirectional microphones (model EM272 XLR, phantom powered) or Telinga (Botarbo, Sweden) stereo Mk2 microphone with parabolic reflector and PRO-handle were used. The microphone output was amplified and digitally recorded in each case using a ZOOM (Tokyo, Japan) f4 recorder. To reduce wind noise, appropriate shields were used in both cases.
5.7 Live recording
For the live mode, the Raspberry Pi version was used. To connect the microphone and headphones, a USB audio adapter (manufactured by ’Plugable’) was used. A FEL Clippy EM272 equipped with 3.5 mm audio jack (powered from USB) was used. Alternatively, a microphone plugged directly into a USB port was used (Richera USB 2.0 Mini Microphone Makio). This is small (2x2 cm including the socket), but less sensitive than the FEL.
Alternatively, for live use, the same setup for recording to file with line-out from the ZOOM (set at mic level) connected into the Raspberry Pi can be used (directly to audio socket or via USB adapter). This method provides a high level of user control over the input prior to analysis.
5.8 Technical setup and optimisation
Frequency range and song phrase durations were determined by birdsong recording analyses. Because the frequencies were in the 1-10 kHz range, high- and low-cut filters near these limits were useful for pre-processing the input. Following the analysis, further filtering is applied to ensure that notes lower than middle C (MIDI note 60, 261 Hz) are not passed to the translation unit. A good recording has a sound amplitude of around 20 dB and a minimum threshold power of about 10 dB, with a default growth (minimum change to report a new note) of 3 dB. The length of notes in birdsong can be difficult to define because the song is complex and changes faster than the human ability to detect. In most cases, a stable time of 20 ms was found to be adequate for analysis. As a default, 256 analysis points were used in each widow, with 256 points between each analysis. If there is a pause in the song of more than 300 ms, the current word is changed to have fewer than four notes and a new word is started. This time limit can be adjusted, but it works well for most birdsongs.
5.9 Other software
Animated text work created with Oleg Cher’s ‘Reedy’ speed reading extension for Google Chrome, which is completely open source and licenced under the GNU GPL v2.
The ‘Read Aloud’ chrome extension (v 1.33.0) was used for the text to voice work with the Microsoft Zira voice.
Screen recordings were made with Blueberry Software’s FlashBack Express 5 recorder/player v5.
Musescore 3 was used to create musical scores from MIDI input (GNU GPLv2 licence).
Acknowledgements
Funding
This work forms part of my practice based PhD in Fine Art at Newcastle University funded by the UK AHRC, Northern Bridge Doctoral Training Partnership.
Supervision
This project was supervised by Richard Grayson (Fine Art), Irene Brown (Fine Art) and Michael Lewis (Philosophy) with support from John Bowers (Culture Lab) – all at Newcastle University.
Other acknowledgements
I am grateful to Mathew Warren for use of the recording of our jamming session and to Kevin Buckley for many stimulating conversations and use of his bird mask.
Appendix
Further technical details and acknowledgements
Solresol dictionary
I used one as originally defined by François Sudre in 1866, Translated by Garrison Osteen in 2011 (see: https://www.sidosi.org/resources#solresol-pages). I have edited this to remove different word options (i.e. to give just one meaning per Solresol word) and remove any punctuation and capitalisation. This lists 2660 words and there are actually 7 + (7x7) + (7x7x7) + (7x7x7x7) = 2800 possible “notewords”, i.e. there are 140 possible that Sudre didn’t define; if there is no match a nearby word is selected.
Software
The software is written using the open source visual programming environment, “Purr Data” by Jonathan Wilkes (https://puredata.info/) – version 2.9.0 . This is based on Pd-l2ork (by Ico Bukvic) which in turn is based on Pd-extended (by Hans-Christoph Steiner). Pd-extended builds on the core Pure Data (version 0.48.0) copyrighted by Miller Puckette and others. Purr data incorporates several other packages which have been used in particular “GEM” – the Graphics Environment for Multimedia (https://puredata.info/downloads/gem/). The pitch analysis is done with the “sigmund.pd” object from Miller Puckette.
Bird song files
Blackbird: https://freesound.org/people/squashy555/sounds/341675/ - CC0 Public Domain
Skylark: https://freesound.org/people/Benboncan/sounds/72752/ - CC BY 3.0
Song Thrush: https://freesound.org/people/inchadney/sounds/117103/ - CC BY 3.0
Otherwise my recordings using Telinga Pro MK2 mic and parabolic reflector with Zoom f4 recorder.
Midi Player / sound font
The Pure Data Midi output is sent to the CoolSoft freeware virtualMIDISynth by Claudio Nicora (https://coolsoft.altervista.org/en/virtualmidisynth). This uses: the Farfisa piano sound font by Riccardo Loi, the FluidR3 GM bank sound font and the GeneralUser sound font from S Christian Collins (see http://www.synthfont.com/soundfonts.html for links to these).
Text font
The Enochian font is freeware created by the Nu Isis Working Group (https://www.scribd.com/document/12599187/NIenoch). An alternative font used was https://www.wfonts.com/font/enochian
Other software
Animated text work made using the “Reedy” speed reading extension for Google Chrome which is fully open source and licenced under GNU GPL v2 written by Oleg Cher.
Screen recordings made using the FlashBack Express 5 recorder/player v5 by Blueberry Software https://www.flashbackrecorder.com/
Musescore 3 (https://musescore.org/en) was used to generate musical scores from MIDI input (GNU GPLv2 licence)
Supervisors
[to be expanded]
Richard Grayson
Irene Brown
Michael Lewis
Also
John Bowers
Funding
This work forms part of my practice based PhD in Fine Art at Newcastle University funded by the UK AHRC, Northern Bridge Doctoral Training Partnership