donghao-zhou/OpenVE-HQ-1M

📌 Dataset Summary

OpenVE-HQ-1M is a high-quality video editing dataset for instruction-guided video editing. It is filtered from OpenVE-3M.

This release does not duplicate video files. The videos are already hosted by OpenVE-3M. OpenVE-HQ-1M provides curated CSV metadata that identifies the filtered high-quality samples.

🗂️ File Structure

The released dataset is organized as follows:

OpenVE-HQ-1M/
├── README.md
└── csv_files/
    ├── background_change.csv
    ├── global_style.csv
    ├── local_add.csv
    ├── local_change.csv
    └── local_remove.csv

• csv_files/: high-quality dataset metadata for the five released categories.
• Video files are not included in this repository. Please download them from OpenVE-3M.

🔗 Using With OpenVE-3M

OpenVE-HQ-1M is intended to be used together with the extracted OpenVE-3M videos. The released training metadata is restricted to samples that can be reliably aligned to public OpenVE-3M video paths. A typical local layout is:

datasets/
├── OpenVE-3M/
│   ├── videos/
│   │   ├── background_change/
│   │   ├── global_style/
│   │   ├── local_add/
│   │   ├── local_change/
│   │   └── local_remove/
│   └── csv_files/
│       └── ...
└── OpenVE-HQ-1M/
    ├── README.md
    └── csv_files/
        ├── background_change.csv
        ├── global_style.csv
        ├── local_add.csv
        ├── local_change.csv
        └── local_remove.csv

The video and original_video fields in OpenVE-HQ-1M use the same relative-path style as OpenVE-3M after extraction:

background_change/edb54dcf7a338c411bc0b8c2157234b3.mp4
global_style/83e667e84c450c3f843ea6bb711a233f.mp4
local_change/b12bd88a8bce0933668592a44b206fe3.mp4

When using this dataset for training, configure the training code with two separate roots:

video root:    <path-to-OpenVE-3M>
metadata root: <path-to-OpenVE-HQ-1M>

The training dataloader should read metadata from OpenVE-HQ-1M and resolve the video and original_video fields using the configured OpenVE-3M video path. For example, if OpenVE-3M videos are available at datasets/OpenVE-3M/videos, then:

background_change/edb54dcf7a338c411bc0b8c2157234b3.mp4

resolves to:

datasets/OpenVE-3M/videos/background_change/edb54dcf7a338c411bc0b8c2157234b3.mp4

🧾 Data Fields

Each row in the category CSV files contains one video editing triplet:

Field	Type	Description
video	path	Relative path to the edited video under the extracted OpenVE-3M videos/ directory.
prompt	string	Editing instruction.
original_video	path	Relative path to the source video under the extracted OpenVE-3M videos/ directory.

The CSV files follow the OpenVE-3M metadata format and use semicolon delimiters:

video;prompt;original_video

📊 Data Size

Category	Samples
background_change	134,822
global_style	170,511
local_add	175,814
local_change	171,561
local_remove	316,870
Total	969,578

🔍 Example Sample

An example entry from csv_files/background_change.csv is shown below:

{
  "video": "background_change/edb54dcf7a338c411bc0b8c2157234b3.mp4",
  "prompt": "Replace the background with a dynamic space station interior where Earth slowly rotates outside the window, soft blinking lights on control panels pulse, and subtle air circulation causes slight movement of hanging objects. The children remain perfectly still in the foreground.",
  "original_video": "background_change/ddf2ac23f1d8ec01091e526714249a13.mp4"
}

The paths above refer to files from the extracted OpenVE-3M video archive, not files stored in this OpenVE-HQ-1M repository.

🎯 Intended Uses

OpenVE-HQ-1M is intended for research and training of instruction-guided video editing systems. Typical use cases include training or fine-tuning video editing models on high-quality OpenVE-3M samples.

This release is intended for research and non-commercial use.

💻 How to Load

You can load the metadata with standard Python tools such as pandas.

from pathlib import Path
import pandas as pd

openve3m_video_root = Path("datasets/OpenVE-3M/videos")
openvehq_root = Path("datasets/OpenVE-HQ-1M")

meta = pd.read_csv(openvehq_root / "csv_files" / "background_change.csv", sep=";")

sample = meta.iloc[0].to_dict()
edited_video = openve3m_video_root / sample["video"]
source_video = openve3m_video_root / sample["original_video"]

print(sample["prompt"])
print(edited_video)
print(source_video)

⚖️ License and Ethics

OpenVE-HQ-1M is derived from OpenVE-3M and is released under CC BY-NC 4.0 for research and non-commercial use.

Users should apply the dataset responsibly and avoid using it for harmful, deceptive, privacy-violating, or otherwise abusive applications. Generated editing results are determined by model behavior and user-provided instructions.