Tag
SVPino highlights Voice Arena AI's Joy, a 100,000-hour dual-channel conversational voice dataset with separate per-speaker tracks, overlapping speech, interruptions, pauses, and backchannels, designed to teach voice models conversational timing.