Proposal of a speaker identification method by adding speech bubbles to avatars in social VR

Identifying the speaker in multi-user Social VR environments remains difficult, especially when multiple users talk simultaneously. This work proposes a speech-bubble-based visual augmentation system that displays keyword- or transcript-based cues above avatars to support speaker-avatar association, which we operationally refer to as speaker identification in this paper. We conducted a controlled exploratory study across 36 conditions (balloon presence, shape, content type, 2D vs. VR) with nine participants. Results show that keyword-based speech bubbles improved speaker-identification performance in the present task relative to both no-bubble and real-time transcription conditions. In contrast, real-time transcription bubbles were associated with lower performance, possibly because of increased visual/textual processing demands and temporal asynchrony in the displayed text. These findings provide preliminary design implications for lightweight visual augmentation in Social VR communication systems and practical UI guidelines for improving clarity and social awareness in immersive multi-speaker environments. We also discuss deployment considerations, including latency, readability constraints, learning effects, topic-related confounds, and limits on generalizability. These findings should be interpreted in light of the controlled experimental setting, the modest and demographically narrow sample, and the exploratory nature of the inferential analyses.