Wishing Well creates an environment where visitor’s interaction shapes the
substance, meaning and character of the artwork – mediated through artificial intelligence. Upon engaging the installation, a participant speaks a wish into a pool of water that only they can hear. Their spoken wish is collected by an artificially intelligent set of software and hardware that begins to process the viewer’s hope. There is a technical document attached to the proposal that details the specific flow and technologies, but in essence, once each wish is recorded it is spoken back in a transformed in a poetic manner by text to speech synthesis in a whispering tone. The individual viewer privately hears this once – their wish then is processed to
become part of a larger collection of all wishes which can be heard through
synthesized voice through directional speakers. The tone of the collection of all wishes is analyzed and affects the processing of the spoken wishes, the ambient sound, and the color and brightness of the light produced above and below the pool.
Wishing Well engages both private and public spaces in the creation of a
simultaneously intimate and collective social experience. The tone of the wishes aggregated by the system is processed anonymously and delivered into this social sculpture via synthesized whispered voice spoken in English. The water pool, or wishing well, is vibrated by a speaker underneath creating a three-dimensional dynamic sound wave image of the actual wish. Virtual lanterns will appear projected with each wish and will launch skyward to become part of the larger collection of dreams.
AI
Speech To Text
IBM Watson’s Speech to Text service accepts an audio file or stream as input and returns a transcription. This service will allow Wishing Well to analyze user speech for its tone and repeat it back using text-to-speech.
Tone Analyzer
IBM Watson’s Tone Analyzer service accepts text as input and returns a tone analysis that detects analytical attitude, inhibition, confidence, hope, sadness and joy across a spectrum. This data will be used to drive the color and intensity of the light of Wishing Well based on the aggregate tone of the wishes.
Polly
Amazon AWS’ Polly output can be modified with SSML tags which can affect the timbre, loudness, and pace of the audio. Wishing Well will whisper wishes back to the wisher.
Cymatics and Spouting Bowls
Cymatics, from the Ancient Greek: κῦμα, signifying, “wave”, is a
subset of modulation physics. Ordinarily the outside of a plate or
vessel is vibrated, and areas of most extreme and least disturbance
are made visible by animating fluid. Various patterns emerge dependent on the amplitude, frequency and shape of the vessel.
The device utilized can be direct, for example, the Chinese spouting bowl, in which copper handles are rubbed and cause the copper base components to vibrate – thus stimulating the water.
Standing vessels started in China. An early form called nao took the
state of a stemmed flagon, mounted with edge upright, and hit
outwardly with a mallet.
Wishing Well User Experience
- The visitor steps into a zone and triggers a proximity sensor that begins recording.
- The user speaks their wish into the microphone in the console.
- The recording stops.
- The audio is converted to text via the speech-to-text AI service.
- The text is sent to the IBM Language Translator and then to Amazon Polly.
- The Language Translator returns the text.
- The text is sent to the Tone Analyzer.
- The Tone Analyzer returns a tone analysis.
- Tone is analysis is sent to Unity and Max to affect light color and sound.
- The tone analysis drives visual and synthesized voice effects based on the mood of the wish.
- The synthesized voice audio is played through a hyper-directional speaker using ultrasound directed toward the visitor so only they can hear privately. http://hypersound.com/pro/products/
- Previous wishes are played at the same time through a randomly selected hyper-directional speaker surrounding the installation. Each visitors wish becomes part of the cloud of wishes and randomly plays anonymously through AI voice.
- The installation goes into an idle / attract state in which visual and auditory effects are influenced by the combined tone of the currently saved wishes.