Conversational speech samples
Browse consented, channel-separated conversations. Each speaker is captured on a separate channel, with manually reviewed transcripts and turn-taking labels where available. Open a sample to read the transcript synced to playback.
English · 2 speakers · spontaneous conversation
audioFull-duplex 2-party English call with disfluency-preserving transcripts. Each speaker is on a separate channel (L / R), so you can mute or solo either speaker in the player.
Korean · 2 speakers · free chat · AI
audioFull-duplex Korean conversation (random chat, topic: AI). Yuna is on the left channel and GleefulAdapter is on the right.
English · 2 speakers · HQ spontaneous conversation
audioFull-duplex 2-party English HQ spontaneous conversation. Zach is on the left channel and Karen is on the right. The provided merged transcript is split by speaker for synchronized playback.
Korean · 2 speakers · free chat · careers
audioFull-duplex Korean conversation (random chat, topic: careers). Yuna is on the left channel and Danmoon is on the right.
Korean · 2 speakers · free chat · time
audioFull-duplex Korean conversation (random chat, topic: time). Danmoon is on the left channel and GleefulAdapter is on the right.
Japanese · 2 speakers · free chat · age
audioFull-duplex Japanese conversation (random chat, topic: age). Ren is on the left channel and Sota is on the right. Speaker metadata is made up because the source package did not include profile metadata.
English · 2 speakers · video conversation · anime & movies
videoFull-duplex 2-party English video conversation (topic: favorite manga, anime, or movie). Video playback requires an allowlisted Google account.
English · 3 speakers · spontaneous conversation
audioFull-duplex 3-party English conversation. Metadata is not included in the source drop, so speaker labels are placeholders while the actual mono stems are used for playback.