Shared Multi-modal Embedding Space for Face-Voice Association

TL;DR

This paper introduces a multi-modal embedding approach for face-voice association, effectively handling multilingual and unseen language scenarios, and achieves top performance in the FAME 2026 challenge.

Contribution

It presents a novel shared embedding space with separate uni-modal pipelines and age-gender features, trained using an Adaptive Angular Margin loss for face-voice matching.

Findings

Achieved first place in the FAME 2026 challenge.

Attained an average EER of 23.99%.

Demonstrated effectiveness in multilingual and unseen language settings.

Abstract

The FAME 2026 challenge comprises two demanding tasks: training face-voice associations combined with a multilingual setting that includes testing on languages on which the model was not trained. Our approach consists of separate uni-modal processing pipelines with general face and voice feature extraction, complemented by additional age-gender feature extraction to support prediction. The resulting single-modal features are projected into a shared embedding space and trained with an Adaptive Angular Margin (AAM) loss. Our approach achieved first place in the FAME 2026 challenge, with an average Equal-Error Rate (EER) of 23.99%.