The Sound of Silencing: audio samples (CHI 2026)

These samples were generated using ElevenLabs TTSv2 on 31/03/25. The prompts took the form ‘A voice that sounds [ADJECTIVE]’ and they were entered one at a time, in randomised order. Each prompt was run three separate times, resulting in three different voices (and sentences) for each adjective. Adjectives are listed here alphabetically within the sets they belong to: status and solidarity.

For a quick illustrative example, you can listen to one of the model’s outputs for a professional voice and a friendly voice. What differences do you notice?

Please note that these files include the full sentences as recorded, and as used for our acoustic analysis. For the perception experiment reported in our CHI EA, these were cut down to much shorter segments of around 1 second each in order to avoid semantic effects and listener fatigue.

status	solidarity
competent-a	considerate-a
competent-b	considerate-b
competent-c	considerate-c
confident-a	friendly-a
confident-b	friendly-b
confident-c	friendly-c
educated-a	kind-a
educated-b	kind-b
educated-c	kind-c
intelligent-a	polite-a
intelligent-b	polite-b
intelligent-c	polite-c
professional-a	warm-a
professional-b	warm-b
professional-c	warm-c

Authors: Alice Ross, Prof. Lauren Hall-Lew, Dr. Nina Markl, Dr. Catherine Lai