---
dataset_info:
  features:
  - name: id
    dtype: string
  - name: audio
    dtype:
      audio:
        sampling_rate: 24000
  - name: emotion_tone
    dtype: string
  - name: speaker_id
    dtype: string
  - name: user_prompt
    dtype: string
  - name: duration
    dtype: float64
  splits:
  - name: train
    num_examples: 15049
---

# sad-batch-ira

Sad Hinglish audio dataset generated using Gemini Live S2S with voice cloning.

- **Samples:** 15,049
- **Emotion:** sad (12k pure sad + 3k sad+sigh)
- **Language:** Hinglish (Hindi-English code-mixed)
- **Audio:** 24kHz mono WAV, 7-30s duration
- **Speaker:** SP_SP010 (voice cloned)