Hinglish, Human Transcripts
Dual-channel Hinglish natural two-speaker conversations with human transcriptions
Not published. Available to buyers under a commercial license.
Sample audio
19:42
Speaker A
Speaker B
Data description
Channel-separated natural two-speaker conversations collected through The Agentic Data Company’s platform. Transcripts are human-verified with comprehensive annotations, such as delivery styles, non-speech verbalism, and mispronunciations. Conversations are between friends, family, partners, and colleagues. The conversations were recorded in diverse, real-world conditions, and microphones and environments vary across sessions. Each speaker’s microphone is captured locally as uncompressed 16-bit PCM at 48 kHz.
The dataset is intended for S2S, TTS, ASR, diarization, and audio understanding.
| Detail | Value |
|---|---|
| Speakers per recording | 2 |
| Channels | Dual-separated (one file per speaker) |
| Audio format | .flac / .wav |
| Sample rate | 48 kHz |
| Metadata format | .json(UTF-8) |
| Transcription type | Human-verified with annotations |