Speaker separated conversational audio.

The dataset primarily consists of natural, spontaneous, and unscripted conversations involving 2 to 5 participants. Each speaker is recorded on a separate, isolated audio channel, ensuring complete voice separation without mixing speakers into a single waveform. Conversations maintain clear turn-taking patterns, with overlapping speech accounting for less than 1% of the recordings.

Request a samplesupport@perit.ai
65,000
Hours collected
18
Languages
13
Countries

English

0:00 / 0:18 · 44.1 kHz · 16-bit

A 66%
B 34%

0.0% overlap· 76% speech· 12.8 turns/min

Transcript

Read along

B · 0:00

A · 0:03

B · 0:12

A · 0:15

Spanish

0:00 / 0:22 · 44.1 kHz · 16-bit

A 66%
B 34%

0.2% overlap· 73% speech· 20.9 turns/min

Transcript

Read along

A · 0:00

B · 0:01

B · 0:07

A · 0:09

B · 0:09

A · 0:10

B · 0:10

A · 0:10

French

0:00 / 0:22 · 44.1 kHz · 16-bit

A 34%
B 67%

0.0% overlap· 73% speech· 13.6 turns/min

Transcript

Read along

A · 0:00

A · 0:04

B · 0:06

A · 0:09

B · 0:11

Arabic

0:00 / 0:16 · 44.1 kHz · 16-bit

A 50%
B 50%

0.0% overlap· 84% speech· 14.7 turns/min

Transcript

Read along

A · 0:00

B · 0:00

A · 0:01

B · 0:09

Chinese

0:00 / 0:36 · 44.1 kHz · 16-bit

A 16%
B 84%

0.0% overlap· 78% speech· 9.8 turns/min

Transcript

Read along

B · 0:00

A · 0:07

B · 0:07

A · 0:08

B · 0:13

B · 0:19

German

0:00 / 0:14 · 44.1 kHz · 16-bit

A 50%
B 50%

0.0% overlap· 82% speech· 8.5 turns/min

Transcript

Read along

A · 0:00

B · 0:07

Portuguese

0:00 / 0:15 · 44.1 kHz · 16-bit

A 54%
B 46%

0.0% overlap· 70% speech· 19.8 turns/min

Transcript

Read along

B · 0:00

A · 0:05

B · 0:06

A · 0:07

A · 0:09

Japanese

0:00 / 0:17 · 44.1 kHz · 16-bit

A 54%
B 46%

0.0% overlap· 76% speech· 27.4 turns/min

Transcript

Read along

B · 0:01

B · 0:04

A · 0:07

A · 0:10

B · 0:10

A · 0:10

B · 0:12

A · 0:13

Korean

0:00 / 0:16 · 44.1 kHz · 16-bit

A 34%
B 66%

0.1% overlap· 72% speech· 28.6 turns/min

Transcript

Read along

A · 0:00

B · 0:00

A · 0:00

B · 0:02

A · 0:07

B · 0:08

A · 0:12

B · 0:14

Italian

0:00 / 0:25 · 44.1 kHz · 16-bit

A 22%
B 78%

0.0% overlap· 76% speech· 9.3 turns/min

Transcript

Read along

A · 0:00

B · 0:03

A · 0:07

B · 0:09

Dutch

0:00 / 0:15 · 44.1 kHz · 16-bit

A 58%
B 42%

0.0% overlap· 87% speech· 7.8 turns/min

Transcript

Read along

B · 0:00

A · 0:06

Polish

0:00 / 0:14 · 44.1 kHz · 16-bit

A 59%
B 41%

0.1% overlap· 71% speech· 47.1 turns/min

Transcript

Read along

B · 0:00

A · 0:00

B · 0:00

A · 0:00

B · 0:01

A · 0:01

B · 0:02

A · 0:03

B · 0:07

A · 0:08

A · 0:10