| --- |
| license: cc-by-4.0 |
| task_categories: |
| - automatic-speech-recognition |
| - text-to-speech |
| language: |
| - vi |
| tags: |
| - audio |
| - tts |
| - vietnamese |
| - conversation |
| - multi-speaker |
| pretty_name: OmniDistil |
| size_categories: |
| - 1K<n<10K |
| dataset_info: |
| features: |
| - name: conversation_id |
| dtype: string |
| - name: turn_index |
| dtype: int32 |
| - name: agent |
| dtype: string |
| - name: voice |
| dtype: string |
| - name: prompt |
| dtype: string |
| - name: transcript |
| dtype: string |
| - name: duration_s |
| dtype: float32 |
| - name: audio |
| dtype: audio |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: data/*.parquet |
| --- |
| |
| ## Dataset columns |
|
|
| | Column | Type | Description | |
| |---|---|---| |
| | `conversation_id` | string | Unique identifier for the conversation | |
| | `turn_index` | int32 | Zero-based turn position within the conversation | |
| | `agent` | string | Agent name | |
| | `voice` | string | TTS voice used for this turn | |
| | `prompt` | string | Full system instruction sent to Gemini Live for this speaker | |
| | `transcript` | string | Text spoken in this turn | |
| | `duration_s` | float32 | Approximate audio duration in seconds | |
| | `audio` | audio | WAV audio (24 kHz mono 16-bit) | |
|
|
