Giving Personality to AI Voices: Towards Parametric Manipulation of First Impressions in Synthetic Voices
KEY WORDS
neurocognition, AI, neuroscience
CITY
Marseille
COUNTRY
France
DETAILS OF THE OFFER
Working place: Institut de Neurosciences de la Timone, Aix-Marseille University
Missions: Offer Description
While synthetic voices now achieve very high levels of naturalness, they remain largely limited to a small number of pre-recorded personas (but see e.g. https://elevenlabs.io/).This project aims to leverage a unique combination of neurocognitive and computer science methods to set the stage for a principled, parametric manipulation of personality impressions in synthetic voices.
Past research has shown that listeners form robust “first impressions” of a speaker’s personality based on brief voice samples, well summarized by a 2-D Trustworthiness-Dominance “Social Voice Space” (SVS). This PhD project will investigate the cerebral response to, and automatic recognition, synthesis and parametric manipulation of voice first impressions in synthetic voices. The DC will: (i) collect personality ratings for a large array of voices in several different languages; (ii) analyze which cerebral response to these voices is predictive of the ratings; (iii) use analysis-synthesis methods relying on a neural speaker-embedding vectors to predict the speaker embedding vectors from the SVS ratings.
The successful DC candidate will be hosted at the Institut de Neurosciences de la Timone at Aix-Marseille University in France under the supervision of Prof. Pascal Belin. The project will be conducted in close co-supervision with Dr Junichi Yamagushi at the Nippon Institute of Informatics, and with Dr Anil Alexander at Oxford Wave Research, and the DC will spend several month at each of these two institutions to help predict the speaker-embedding vectors from verbal descriptions of speaker traits and explore potential use-cases for expressive, controllable speech synthesis in commercial and law enforcement applications.
Qualifications:
Background: Academic background matching with the research project (Computer science).
Degree: At the date of the recruitment, candidates must be in possession of a Master’s degree in a relevant field and should not have been already awarded a doctoral degree. Applicants must have submitted their master’s or equivalent thesis in relevant fields. The diploma/certificate must be forwarded as soon as it is received. Researchers who have successfully defended their doctoral thesis but who have not yet formally been awarded the doctoral degree will not be considered eligible.
Mobility: Recruited researchers can be of any nationality but must not have resided or carried out their main activity (work, studies, etc.) in the country of their host organization for more than 12 months in the 3 years immediately prior to their recruitment. Short stays, such as holidays, are not taken into account.
English language: Candidates must demonstrate their ability to understand and express themselves sufficiently in both written and spoken English, in order to fully benefit from the network training.
