Holographic calling systems aim to create the perceptual illusion that a remote interlocutor is physically present by combining augmented reality (AR) visualizations with spatial audio rendering. A key cIRCAM:Galleryenge in such systems is achieving audiovisual coherence when room-acoustic properties must be inferred, since this can lead to greater inaccuracies than when they are estimated from measurements. In this study, we investigate perceptual tolerance to mismatches in the direct-to-reverberant energy ratio (DRR), a critical cue for auditory distance perception. We conducted a listening experiment in which participants judged whether the audio and the visual presentation of a stimulus were coherent using a yes/no task. Stimuli included both a volumetric capture of a speaking human avatar, rendered at the correct direct sound level, and a loudspeaker reproducing wideband noise. For the loudspeaker, level roving was introduced to assess the influence of intensity cues on listener decisions. Results show that audiovisual coherence is maintained for half of the presentations within a DRR mismatch range of approximately 3 dB for too dry and 4.3 dB for too reverberant renderings. Within the limited number of participants included in the study, no evidence for significant differences was found between speaking avatars and loudspeakers reproducing noise, nor between different ranges of level roving in the loudspeaker condition. Nevertheless, the findings help to understand the consequences of DRR estimation mismatches for holographic AR calling experiences.
