---
description: "Load audio datasets from various sources including FLEURS, custom manifests, and local files"
categories: ["workflows"]
tags: ["data-loading", "audio-manifests", "fleurs", "local-files", "batch-processing"]
personas: ["data-scientist-focused", "mle-focused"]
difficulty: "beginner"
content_type: "workflow"
modality: "audio-only"
---

# Load Audio Data

Import audio datasets from various sources into NeMo Curator's audio processing pipeline. Audio data loading supports manifest files, direct file paths, and automated dataset downloads.

## How it Works

Audio data loading in NeMo Curator centers around the `AudioTask` data structure, which contains:

- **Audio file paths**: References to audio files (.wav, .mp3, .flac, and so on)
- **Transcriptions**: Ground truth or reference text for speech content
- **Metadata**: Duration, language, speaker information, and quality metrics

The loading process validates audio file existence and formats data for downstream ASR inference and quality assessment stages.

---

## Loading Methods

Choose the appropriate loading method based on your data source and format:

<Cards>

<Card title="FLEURS Dataset" href="/curate-audio/load-data/fleurs-dataset">
Automated download and processing of the multilingual FLEURS speech dataset
automated
multilingual
102-languages
</Card>

<Card title="Custom Manifests" href="/curate-audio/load-data/custom-manifests">
Create and load custom audio manifests with file paths and transcriptions
jsonl
tsv
custom-format
</Card>

<Card title="Local Files" href="/curate-audio/load-data/local-files">
Load audio files directly from local directories and file systems
local-storage
batch-processing
file-discovery
</Card>

</Cards>