Rendered spoken audio for the DuplexGen turn-taking dialogues — the exact
set of clips used to fine-tune the full-duplex model (PP-DG) in
DuplexGen: Adaptive Synthesis of Human–AI Turn-Taking Dialogues.
Each clip is a full render of one generated dialogue variation: the mixed
two-speaker dialogue audio, the isolated per-turn utterances, and the inserted
backchannel clips, plus per-clip metadata. Audio is synthesized with
Chatterbox TTS; the dialogue
text it… See the full description on the dataset page:
https://huggingface.co/datasets/DuplexGen/duplexgen-spoken.
