---
license: mit
task_categories:
- text-to-speech
language:
- en
tags:
- audio
- tts
- synthetic
- emotion
- kion
pretty_name: KionTTS Dataset
size_categories:
- 10K<n<100K
dataset_info:
  features:
    - name: id
      dtype: string
    - name: tier
      dtype: string
    - name: text
      dtype: string
    - name: audio
      dtype: audio
    - name: emotions
      dtype: string
    - name: styles
      dtype: string
configs:
  - config_name: default
    data_files:
      - split: tier1
        path: data/tier1/*.parquet
      - split: tier2
        path: data/tier2/*.parquet
      - split: tier3
        path: data/tier3/*.parquet
---

# KionTTS Dataset

Synthetic TTS training dataset for **Kion** (Kognetive Intelligence Orchestration Node),
generated with IndexTTS2 using emotion and style conditioning.

## Dataset Structure

| Field | Type | Description |
|-------|------|-------------|
| `id` | string | Unique 8-char hex ID |
| `tier` | string | `tier1` (regenerated fixes), `tier2` (spot-check), `tier3` (clean) |
| `text` | string | Full text with inline emotion tags e.g. `[happy=0.7] Hello!` |
| `audio` | Audio | WAV audio, 24kHz mono |
| `emotions` | string | JSON dict of emotion labels and intensities |
| `styles` | string | JSON dict of style labels and intensities |

## Tiers

- **Tier 1** (3,323 files) — Regenerated with corrected emo_alpha to fix reference speaker leakage
- **Tier 2** (933 files) — Medium-risk; recommend spot-checking before training use
- **Tier 3** (14,517 files) — Clean; safe for direct training use

## Loading the dataset

```python
from datasets import load_dataset

# Load clean tier only
ds = load_dataset("nate0001/KionTTS", split="tier3")

# Load all tiers
ds = load_dataset("nate0001/KionTTS")
```