otoSpeech datasets
All datasets are full-duplex and recorded with speaker consent.
Casual Language Speakers Media License Actions otoSpeech 1,125h English 2 Audio Commercial English casual conversation with a separate channel for each speaker.
Sample rate 48 kHz
Sessions 3,878
Includes Speaker profiles About English datasets
otoSpeech 280hEnglish 2 Audio CC BY 4.0 Raw English conversation recorded in varied real-world conditions. Microphones and rooms differ across sessions.
Sample rate 48 kHz
Includes Stereo FLAC, session metadata, speaker profiles, redaction intervals, and participant surveys About English datasets
otoSpeech 141hEnglish 2 Audio CC BY 4.0 Conversations selected from the 280h release by human review, with noise reduction and speech enhancement. It also adds conversations recorded after that release.
Sample rate 44.1 kHz
Includes Stereo FLAC, session metadata, speaker profiles, redaction intervals, and participant surveys About English datasets
otoSpeech Turn 104hTurn-taking English 2 Audio oto Speech License 420 conversations annotated by hand for turn-taking. This is the training set for TurnBench, the open benchmark from Sesame.
Sample rate 48 kHz
Labels 17 event labels per speaker, including backchannels, interruptions, floor holds, fillers, and laughter
Includes One WAV per speaker, a stereo version, labeled SRT transcripts, and session metadata
License Non-commercial research. Commercial use needs a separate written license. About the TurnBench training set
otoSpeech VideoEnglish 2 Audio Video On request English conversation with a camera view of each speaker, synced to two-channel audio. Faces are mosaicked in the sample to protect the speakers' privacy. The licensed data has no mosaic.
Sample rate 48 kHz
Includes One camera video per speaker, stereo audio, and per-speaker transcripts About the video dataset
otoSpeech Face-to-Face English 2-5 Audio Video On request English conversation recorded in person, with two to five people. Sessions are casual or deliberative.
Room 23.1 m² floor area, 46.27 m³ volume, 2.3 m ceiling
Seating Four speaker stations in alternating positions
Microphones Sony C-38B and a Sony lavalier (ECM-77B or ECM-44B) at every station. Shure SM7B and Sennheiser MKE600 at stations 1 to 3. Sennheiser MKH416 at station 3.
Room mic AKG C414 XLII near the center of the room
Channels 16 channels, 48 kHz, 24-bit WAV
Note This is the main setup. Some later sessions used different microphones. About multi-speaker datasets
otoSpeech 3-SpeakerEnglish 3 Audio On request English conversation among three people, with a separate recording for each speaker.
Sample rate 48 kHz
Includes One mono audio file per speaker About multi-speaker datasets
otoSpeech KoreanKorean 2 Audio On request Korean free conversation on a given topic, with one speaker on each channel.
Sample rate 44.1 kHz
Includes Stereo audio and per-speaker ASR transcripts About the Korean dataset
otoSpeech JapaneseJapanese 2 Audio On request Japanese free conversation on a given topic, with one speaker on each channel.
Sample rate 48 kHz
Includes Stereo audio and per-speaker transcripts About Japanese datasets
Task-oriented otoSpeech Task 20hEnglish 2 Audio Task log CC BY 4.0 58 sessions across seven collaborative tasks. What each speaker saw and did is recorded on the same timeline as the audio.
Sample rate 48 kHz
Includes Stereo FLAC, 11,025 timestamped events, task stimuli, reference answers, and participant artifacts
Not included Transcripts About otoSpeech Task
otoSpeech Task v2 17hEnglish 2 Audio Task log CC BY 4.0 47 sessions across eleven collaborative tasks, recorded after the 20h release. None of its sessions repeat here.
Sample rate 48 kHz
Includes Stereo FLAC, 8,301 timestamped events, task stimuli, and ten participant artifacts About otoSpeech Task
otoSpeech EgocentricEnglish 2-4 Audio Video On request English conversation with first-person video, as two to four people fold towels, tidy everyday objects, or play Jenga.
Audio A close microphone for every person, plus left-ear and right-ear microphones on the main worker
Video Head-mounted cameras on the main worker and one other participant, plus wide and close fixed views About full-duplex egocentric data
Role-play otoSpeech Role-PlayJapanese 2 Audio On request Japanese two-speaker role-play in set scenarios, from shop visits to business talks. The speakers use their own words.
Sample rate 48 kHz About Japanese datasets
Scripted otoSpeech ScriptedScript Japanese 2 Audio On request Japanese two-speaker conversation read from scripts, with each speaker on their own channel. Every recording comes with its script and a timed transcript.
Sample rate 48 kHz About Japanese datasets
Need a dataset that isn't listed? We record and annotate to your specification. Request Looking for a specific kind of data? See English , Japanese , Korean , video , multi-speaker , and egocentric conversation.
Hugging Face