Dataset Card for otoSpeech-full-duplex-processed-141h: Full-Duplex Conversational Speech Dataset

Dataset Summary

otoSpeech-full-duplex-processed-141h is a full-duplex, two-speaker conversational speech dataset. It is derived from otoSpeech-full-duplex-280h and has been curated and processed as follows:

• Selected high-quality conversations based on human reviews.
• Applied noise reduction and speech enhancement to improve audio quality.
• Added new samples collected after the release of otoSpeech-full-duplex-280h.

Each sample includes 44.1 kHz, channel-separated audio, along with session metadata, speaker profiles, redaction intervals, and participant-written surveys.

The conversations were recorded in diverse, real-world conditions. Natural conversational phenomena—such as overlaps, interruptions, and laughter—are preserved, and microphone and room conditions vary across sessions.

This release contains approximately 141 hours of processed English conversational audio.

Note: The original raw, unfiltered version (otoSpeech-full-duplex-280h) is available at:: https://huggingface.co/datasets/otoearth/otoSpeech-full-duplex-280h

Languages

• English (en)

How to use

from datasets import load_dataset

ds = load_dataset("otoearth/otoSpeech-full-duplex-processed-141h", split="train")
print(ds[0].keys())
print(ds[0]["flac"])

Note: This dataset is packaged as WebDataset (tar shards). In each shard, each sample includes flac and json files, plus WebDataset metadata keys such as __key__ and __url__.

---

Dataset Details

Please refer to the otoSpeech-full-duplex-280h repository for details: https://huggingface.co/datasets/otoearth/otoSpeech-full-duplex-280h

Contact

The data released this time is a part of the dataset we possess. If you would like the remaining dataset or a dataset with special requirements, please contact us at the address below.

Website: https://oto.earth Email: consome@oto.earth

We can acquire various types of data according to your requests. For example:

• Full-duplex conversation audio with 3 or more speakers
• Full-duplex conversation audio with specific accents
• Full-duplex conversation audio during gaming
• Full-duplex conversation audio in specific situations (e.g., customer support, doctor-patient conversations)

Citation Information

@misc{otoSpeech-full-duplex-processed-141h,
  title        = {otoSpeech-full-duplex-processed-141h: Full-Duplex Conversational Speech Dataset},
  author       = {otoearth},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/otoearth/otoSpeech-full-duplex-processed-141h}},
  note         = {License: CC BY 4.0}
}