Back to downloads
Research Collaboration

Access BVD-RAW

BVD-RAW contains 80M raw videos totalling 10M hours and is available to research collaborators.
Follow the three steps below to request and receive the data.

1

Accept the Terms of Use

Submit a short request so we can verify your intended use

BVD-RAW is available exclusively for academic and non-commercial research. To request access, please fill in the form below. We ask for your institutional affiliation, a brief description of your research, and to accept the Terms of Use. Requests are reviewed on a rolling basis and you will receive a confirmation once access is granted.

2

Create a Transfer Manifest

Specify which videos you need and where they should be sent

Data is transferred directly to your own S3-compatible storage bucket via GRASS. You will need to create a JSONL manifest file that declares your bucket credentials and lists each video you want. Video IDs can be found in BVD-RAW-Index on Hugging Face.

Each JSON object in the manifest must be compacted to a single line. The examples below show pretty-printed versions for readability; the final example shows the correctly formatted JSONL file.

Line 1 — S3 credentials

JSON
{
  "endpoint":    "{your-s3-compatible-endpoint}",
  "accessKeyId": "{your-access-key-id}",
  "secretKey":   "{your-secret-key}"
}

Subsequent lines — one copy instruction per video

Each video is identified by its YouTube-style ID (e.g. --Uuf_zdsTk). Every entry requests the .mp4, .mp3, and .info.json files for that video and specifies your destination bucket.

JSON — pretty-printed for reference
{
  "id": "{video-id}",
  "do": "copy-v2",
  "doArgs": {
    "manifest": [
      {
        "name": "{video-id}/{video-id}.mp4",
        "bucket": {
          "name": "laion-bucket",
          "credentials": { "endpoint": "stor.grass-storage.io" }
        }
      },
      {
        "name": "{video-id}/{video-id}.mp3",
        "bucket": {
          "name": "laion-bucket",
          "credentials": { "endpoint": "stor.grass-storage.io" }
        }
      },
      {
        "name": "{video-id}/{video-id}.info.json",
        "bucket": {
          "name": "laion-bucket",
          "credentials": { "endpoint": "stor.grass-storage.io" }
        }
      }
    ],
    "destinations": [
      {
        "bucket": {
          "name": "{your-bucket-name}",
          "credentials": {
            "endpoint":    "{your-endpoint}",
            "accessKeyId": "{your-access-key-id}"
          },
          "options": { "includeFileNameParent": true }
        }
      }
    ]
  }
}

Full example — ready-to-use JSONL file

Replace the example credentials and bucket details with your own before submitting.
JSONL — laion-manifest.jsonl
{"endpoint":"s3.storage.com","accessKeyId":"3e9375da808b3202","secretKey":"63144cc9461e335b"}
{"id":"--Uuf_zdsTk","do":"copy-v2","doArgs":{"manifest":[{"name":"--Uuf_zdsTk/--Uuf_zdsTk.mp4","bucket":{"name":"laion-bucket","credentials":{"endpoint":"stor.grass-storage.io"}}},{"name":"--Uuf_zdsTk/--Uuf_zdsTk.mp3","bucket":{"name":"laion-bucket","credentials":{"endpoint":"stor.grass-storage.io"}}},{"name":"--Uuf_zdsTk/--Uuf_zdsTk.info.json","bucket":{"name":"laion-bucket","credentials":{"endpoint":"stor.grass-storage.io"}}}],"destinations":[{"bucket":{"name":"laion","credentials":{"endpoint":"s3.storage.com","accessKeyId":"3e9375da808b3202"},"options":{"includeFileNameParent":true}}}]}}
{"id":"--U3b95ZhXE","do":"copy-v2","doArgs":{"manifest":[{"name":"--U3b95ZhXE/--U3b95ZhXE.mp4","bucket":{"name":"laion-bucket","credentials":{"endpoint":"stor.grass-storage.io"}}},{"name":"--U3b95ZhXE/--U3b95ZhXE.mp3","bucket":{"name":"laion-bucket","credentials":{"endpoint":"stor.grass-storage.io"}}},{"name":"--U3b95ZhXE/--U3b95ZhXE.info.json","bucket":{"name":"laion-bucket","credentials":{"endpoint":"stor.grass-storage.io"}}}],"destinations":[{"bucket":{"name":"laion","credentials":{"endpoint":"s3.storage.com","accessKeyId":"3e9375da808b3202"},"options":{"includeFileNameParent":true}}}]}}
3

Submit Your Manifest to GRASS

GRASS will process your manifest and transfer the files directly to your bucket

Once your manifest file is ready, submit it to GRASS for processing. GRASS will read each entry line-by-line, copy the requested files from the LAION storage pool, and deliver them to the S3 bucket you specified. Large manifests are processed in batches; you will be notified once your transfer is complete.

Submission instructions and the GRASS endpoint will be provided after your access request (Step 1) is approved.
While access to the data is free for research institutions, GRASS may request that you cover egress fees of up to $2.5/TB.