Skip to content

Dataset · otoearth

English full-duplex dialogues (141 hours)

otoearth/otoSpeech-full-duplex-processed-141h

Processed audio recordings of live conversations between two English speakers for training voice models.

What's inside

Contains English audio recordings structured into fewer than a thousand records with a total duration of 141 hours. This is a cleaned and processed subset of the larger otoSpeech-full-duplex-280h corpus, containing natural conversations between two people in full-duplex mode.

What you can do with it

  • Training conversational audio-to-audio models capable of interrupting and listening simultaneously
  • Training speech recognition systems under conditions of overlapping speaker voices
Tasks
Audio → audio
Languages
English
Size
up to 1K rows
Downloads this month
1.4K
Likes
36
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("otoearth/otoSpeech-full-duplex-processed-141h")