Conversational Audio
Overlapping Speech
Real conversations don't take turns politely: speakers interrupt, agree mid-sentence and talk over each other. Because each speaker sits on their own channel, both sides of every overlap stay fully intelligible. Overlapping words are wrapped in <overlap> spans, and back-channels like [mm] and [hmm] are tagged inline.
125Koverlap spans
Request samples Specification
- Markup
- <overlap>...</overlap> inline spans
- Back-channels
- [mm], [hmm], [um], [uh]
- Separation
- Isolated channel per speaker
- Timing
- Turn start / end in seconds
Delivery layout
- ampllab-speech-to-speech-batch-<timestamp>/
- session_<id>/
- scene_<n>/
- take-<n>/
- <speaker-a>.flac
- <speaker-b>.flac
- merged-<take>.flac
- annotation-<id>.json
Inline markup
- <overlap>...</overlap>
- [mm]
- [hmm]
- [um]
- [uh]
- [unintelligible]
take-<n>/annotation-<id>.jsonIllustrative values
{ "script": { "category": "Banking", "sub_category": "Card Servicing" }, "scene": { "id": "scene_2", "name": "Declined card at checkout" }, "take": { "uid": "7f3c9e21-...", "seq": 4, "duration": 287.42 }, "speakers": { "a41d07c2-...": { "role": "Priya (Card Services Agent)" }, "e9b2f6a8-...": { "role": "Daniel (Customer)" } }, "transcription": [ { "speaker": "e9b2f6a8-...", "start_time": 3.05, "end_time": 7.48, "transcript": "[sigh] Hi, my card got declined twice <overlap>this morning</overlap>.", "emotion": "Frustrated", "languages": "English" }, { "speaker": "a41d07c2-...", "start_time": 7.1, "end_time": 11.86, "transcript": "<overlap>Oh no,</overlap> let me take a look [mm]. Can you confirm the last four digits?", "emotion": "Calm", "languages": "English" } ]}Sample pack
A curated set of takes from this dataset, with audio and annotation files in the delivery format.
More in Conversational Audio
Dual-Channel Conversations
Two-speaker conversations with each voice on its own isolated, time-synchronized channel, plus a merged mix of the take.
179 hours · 369 sessions
Domain Role-Play
Scripted agent-customer conversations across banking, insurance, travel, support, sales and negotiation, performed with natural pacing.
9 domains · 44 scenarios