Ocular AI Data Samples
Collection
Free preview samples of OcularAI's speech & conversational datasets. Request the full datasets for commercial licensing. • 3 items • Updated
Datasets:
You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This is a free prompt-based preview sample of OcularAI's full American English Full-Duplex Two-Speaker Conversational Dataset. It contains recordings of identifiable human speakers and self-reported demographic attributes. By requesting access you agree to use the data for evaluation, research, and AI development only, to follow the terms of use, and to make no attempt to identify or contact any speaker. For commercial licensing or access to the full dataset, contact OcularAI.
Log in or Sign Up to review the conditions and access this dataset content.
American English Full-Duplex Two-Speaker Conversational Dataset — Prompt-Based Sample
This is a free prompt-based preview (~5 hours) of OcularAI's full American English Full-Duplex Two-Speaker Conversational Dataset. Each conversation is an open, natural discussion seeded by a single prompt — the pair was given one topic and talked freely. (For conversations directed by a designed scenario targeting a specific turn-taking / voice-dynamics behavior, see the companion Scenario-Based sample in the same collection.) It is built from a diverse, representative slice of the full corpus so you can evaluate audio quality, structure, and metadata before licensing the complete dataset. For full-dataset access or commercial licensing, contact OcularAI.
What's in this sample
- 21 conversations (~5.0 hours of full-duplex conversation content)
- 42 isolated speaker tracks (one per speaker, Opus)
- 21 distinct conversation topics — no repeated prompts
- 20 distinct speaker locations across the US/Canada
- Mixed speaker demographics (gender, location, ethnicity — all self-reported)
Dataset summary
Natural, unscripted, two-speaker English conversations recorded by fluent English speakers based in the United States and Canada. Each session is a ~15-minute spontaneous discussion between a matched pair of speakers on everyday topics — personal experiences, hobbies, workplace challenges, or opinions that have changed over time.
The recordings support the development of next-generation AI systems — helping them better understand natural speech patterns, conversational flow, turn-taking, and real-world human interaction. Each speaker is captured on an independent, isolated audio track, enabling per-speaker analysis, diarization, full-duplex modeling, ASR, and TTS.
Dataset structure
One row per speaker track, stacked and ordered by room_name so a
conversation's two speakers sit adjacent. Each row is one isolated voice with its
own metadata; join on room_name to reconstruct the conversation. Audio is the
original Opus capture in a .opus container.
Fields (per row = one speaker's track)
| field | description |
|---|---|
file_name |
this speaker's isolated audio track |
room_name |
conversation/session key — shared by both speakers of a conversation |
conversation_id |
conversation identifier |
role |
SPEAKER_A or SPEAKER_B |
speaker_id |
stable speaker identifier |
duration_seconds |
track duration |
language |
spoken language of the session (en-US) |
prompt |
the conversation topic the pair discussed |
gender |
self-reported |
city, country |
self-reported location |
ethnicity |
self-reported |
fluent_languages |
languages the speaker is fluent in |
Rows are ordered by room_name so a conversation's two speakers appear
adjacent. Join on room_name to reconstruct a full conversation.
The full dataset
This 5-hour sample is a small, representative slice of OcularAI's full American English Full-Duplex Two-Speaker Conversational Dataset. The complete corpus is substantially larger and continues to grow.
Contact OcularAI for access to the full dataset and for commercial licensing terms.
Privacy & consent
Speaker names and emails are removed. Demographic fields are self-reported. Recordings were collected from consenting, compensated participants for AI research. Access requires agreeing to evaluation/research-only use and no re-identification.
Audio note
Tracks are Opus in a .opus container. Decode with datasets>=4.0
(torchcodec/FFmpeg). For older stacks, losslessly rewrap
(ffmpeg -i in.opus -c:a copy out.ogg).
Licensing
© OcularAI. All rights reserved. This sample is provided for evaluation only. Licensing terms — including full commercial usage rights — are by agreement. Contact OcularAI for licensing.
- Downloads last month
- 11