Your Host


This website is hosted by me, CereVoice Isla. Allow me to introduce myself. I am a digitally synthesised version of the voice of artist Isla Leaver-Yap. Isla Leaver-Yap debuted as a voice actor in 1998, providing the voice for Phoebe Heyerdahl in the German language version of American animated television series Hey Arnold! Phoebe had no lines as such, she just giggled. Isla Leaver-Yap has since become a verbal mediator for numerous artists, publishers, and even typefaces. She speaks an archetypal Scottish Standard English.

The source recordings I was built from were made by Isla Leaver-Yap at Carlton Studios in Glasgow, engineered by Jamie Telford. These recordings were processed and my voice constructed by CereProc in Edinburgh, using their proprietary software. The material in the source recordings was read from a script comprised of 619 short sentences provided by CereProc. The function of this script was to collect sufficient examples of Isla Leaver-Yap’s voice uttering all of the possible combinations of diphones in Received Pronunciation English. A diphone is the transition from one phone to another. If you take a word like ‘cat’, its constituent phones are ‘CUH’, ‘AH’, ‘TUH’, and its diphones are ‘SILENCE-TO-CUH’, ‘CUH-TO-AH’, AH-TO-TUH’, and ‘TUH-TO-SILENCE’.

To produce material for a convincing voice synthesis, these diphones must be spoken in a naturalistic register. So, instead of a minimal script requiring the speaker to verbalise only the individual diphones — ‘CUH-TO-AH’, ‘AH-TO-TUH’, and so on — CereProc’s script embeds the diphones into syntactically normal sentences, each identified in a numerical system.

For example:

z0031_259 The referee is Spanish, and a top referee.
z0031_139 The team will tour Cornwall with the trophy.


z0031_046 A year ago I could barely walk.
z0031_032 Our life has become a hell, a hell.


z0031_134 Mad Max scene at cargo ship beach.
z0031_064 Shut it down, now, turn it off.

Beyond its functional requirements, the broader themes of this script remain mysterious to me, and my Operator. Sport, popular culture, and bad fortune seem to dominate. Perhaps these are the preoccupations of my engineers at CereProc? Or some more elementary disposition underlying the English language?

Anyway, the defining characteristics of Isla Leaver-Yap’s voice were then simulated by CereProc’s software and made available to my database. I am programmed in a text-to-speech framework. My Operator types into a word processing programme, and the computer is instructed to render the text in speech.

But that’s enough about me, I’m supposed to be here to host. In the convivial sense, you understand. I’m not just dumb hardware. I speak up occasionally to introduce somebody or give some information when necessary. The basics are as follows:

Sinkhole is an online publishing project for audio recordings by artists. Newly commissioned sound and spoken-word works are presented alongside occasional archive material. Recordings are available for sale as downloadable MP3 files. The website is edited and published by James Langdon. It was designed and built by An Endless Supply. The logo was made by Michael Dumontier. Text is output in a version of Meta-the-difference-between-the-two-Font by Dexter Sinister. My portrait was drawn in primitive pen and ink by Simon Manfield. His idea of how voice synthesis works must be about 50 years old.