---
pretty_name: "Align JSON Audio Dataset"
task_categories:
- automatic-speech-recognition
tags:
- audio
- asr
- hungarian
---

# Align JSON Audio Dataset

Dataset generated from manifest rows where audio path is resolved from `align_json` (removing `_align.json` and finding matching audio extension), while the `align_json` column stores the alignment file content.

## Summary

- Repo: `sarpba/data_test3`
- Rows in train split: **181259**
- Skipped rows: **0**
- Duration total: **353.97 h**
- Duration mean: **7.03 s**
- Duration median: **4.59 s**
- Duration p95: **21.16 s**
- Duration min/max: **1.00 s / 30.00 s**
- Non-empty `text`: **181259**
- Non-empty `phonetic`: **181259**
- Total transcript characters: **18582604**
- Total transcript words: **2815819**

## Schema

- `id` (string)
- `audio` (Audio)
- `text` (string)
- `phonetic` (string)
- `duration` (float64)
- `align_json` (string, JSON content from *_align.json)

## Audio Extension Distribution

| Extension | Count |
|---|---:|
| `.mp3` | 119516 |
| `.wav` | 61743 |

## Skipped Row Reasons

| Reason | Count |
|---|---:|
| `none` | 0 |