Joy Dataset Released: 100K Hours of Dual-Channel Human Dialogue

The voicearena team released Joy, a dual-channel conversational dataset featuring 100,000 hours of unscripted two-person dialogues with separate audio tracks per speaker, spanning 10 languages and 50,000 speakers. It aims to teach voice models when to take turns or interject in natural conversation.

2026-10-30 ~ 2026-10-30 · 2 related posts

• Joy dataset: 100,000 hours of unscripted two-person calls to teach voice models real timing — Scobleizer · 2026-10-30

1 near-duplicate retellings: Aiden_Tech_Ai