---
license: apache-2.0
task_categories:
  - summarization
language:
  - en
---

# 📚 Clip-CC Dataset

**Clip-CC** is a high-quality dataset of 200 short educational videos (1 minute 30 seconds each) paired with human-written summaries. It’s designed for tasks like video summarization, video-language alignment, and multimodal learning.

---

## 🧾 Dataset Summary

- 🎞️ 200 videos: Named `1.mp4` to `200.mp4`
- 📝 Each video has a concise human-written summary
- 📂 File references stored in a JSONL file (`metadata.jsonl`) with paths to each video

---

## 💡 Use Cases

- Video summarization and caption generation
- Vision-Language alignment
- Video QA and downstream educational AI tasks
- Fine-tuning multimodal models (e.g., Flamingo, Video-BERT, LLaVA)

---

## 📦 Dataset Structure

Each entry has:

| Field      | Description                              |
|------------|------------------------------------------|
| `id`       | Unique ID, e.g., `001`                   |
| `file_name`| Relative path to video file              |
| `summary`  | Human-written summary of video content   |

### 🔍 Example

```json
{
  "id": "001",
  "file_name": "clips/1.mp4",
  "summary": "An introduction to gravitational force using animations."
}