---
license: cc-by-4.0
task_categories:
  - text-to-audio
  - audio-classification
language:
  - en
tags:
  - audio
  - video
  - captions
  - scenes
  - urls
  - laion
size_categories:
  - 10M<n<100M
configs:
  - config_name: data
    default: true
    data_files:
      - split: train
        path: data/*.parquet
---

# LAION-BVD — 10M Audio Clip URLs

This repository contains the metadata and captions for **~10 million audio clips** randomly sampled
from [BVD-V-55M](https://huggingface.co/datasets/laion/BVD-V-55M) for large-scale audio pre-training.

The audio itself is **not** included in this repository — every clip is described by the URL of
its source video plus the `start_time`/`end_time` offsets needed to reproduce it. The
corresponding clip files are available in the gated
[laion/BVD-A-10M](https://huggingface.co/datasets/laion/BVD-A-10M) repository.

## Dataset structure

One row per audio clip (~10M rows).

| Column | Type | Description |
|---|---|---|
| `id` | string | Source video ID (e.g. YouTube ID) |
| `title` | string | Title of the source video |
| `description` | string | Description of the source video |
| `upload_date` | string | Upload date of the source video (YYYYMMDD) |
| `channel` | string | Channel name |
| `webpage_url` | string | URL of the source video |
| `caption` | string | Caption describing the clip |
| `clip_id` | int64 | Identifier of the clip within its source video |
| `language` | string | Detected language of the caption |
| `scene_index` | int64 | Index of this clip within the source video |
| `start_time` | float64 | Clip start time in seconds |
| `end_time` | float64 | Clip end time in seconds |
| `duration` | float64 | Clip duration in seconds |
| `tags` | string | Tags of the source video |
| `category` | string | Category of the source video |
| `recaption_prompt` | string | Prompt used to generate `caption` |
| `recaption_model` | string | Model used to generate `caption` |

## Relation to LAION-BVD

| Release | Description |
|---|---|
| [laion/BVD-URLs](https://huggingface.co/datasets/laion/BVD-URLs) | 1.3B video URLs extracted from CommonCrawl |
| [laion/BVD-RAW-Index](https://huggingface.co/datasets/laion/BVD-RAW-Index) | File index of the ~80M downloaded raw videos (media by request) |
| [laion/BVD-V-55M](https://huggingface.co/datasets/laion/BVD-V-55M) / [-URLs](https://huggingface.co/datasets/laion/BVD-V-55M-URLs) | 55M scene-level video clips with captions and timestamps |
| [laion/BVD-A-10M](https://huggingface.co/datasets/laion/BVD-A-10M) / **-URLs** (this dataset) | 10M audio clips randomly sampled from BVD-V-55M |
| [laion/BVD-A-1.7M](https://huggingface.co/datasets/laion/BVD-A-1.7M) / [-URLs](https://huggingface.co/datasets/laion/BVD-A-1.7M-URLs) | 1.7M audio clips from BVD-V-55M with unique source videos |
| [laion/BVD-I-300M](https://huggingface.co/datasets/laion/BVD-I-300M) / [-URLs](https://huggingface.co/datasets/laion/BVD-I-300M-URLs) | 300M keyframes sampled from BVD-RAW |

BVD-RAW itself (the raw pool of 80M videos totalling 10M hours) is not distributed through
Hugging Face. The gated subsets and BVD-RAW are available to academic and non-commercial
researchers through a single central [access request form](https://forms.gle/5tXPS4gmBwSGYYbV8); the
[BVD-RAW access guide](https://projects.laion.ai/bvd/bvd-raw-access.html) describes the raw-corpus transfer procedure.

## Links

- Project page: <https://projects.laion.ai/bvd/>
- Paper: <https://arxiv.org/abs/2608.24845>
- Code: <https://github.com/LAION-AI/bvd>
- Terms of Use: [bvd_terms_of_use.pdf](https://projects.laion.ai/bvd/assets/bvd_terms_of_use.pdf)
- Access request form: <https://forms.gle/5tXPS4gmBwSGYYbV8>