README.md

# Streamo
<h1 align="center">Streaming Video Instruction Tuning</h1>
<p align="center"><i>A real-time streaming video LLM that serves as a general-purpose interactive assistant.</i></p>

<p align="center">
         📑 <a href="https://arxiv.org/abs/2512.21334">Paper</a>  &nbsp|&nbsp  🌐 <a href="https://jiaerxia.github.io/Streamo/">Web</a>  &nbsp|&nbsp 🤗 <a href="https://huggingface.co/datasets/maifoundations/Streamo-Instruct-465K">Huggingface</a>
</p>

This is the official implementation of the paper 'Streaming Video Instruction Tuning'.


# News📰
* **`[2026/2/27]`:**🎉**Our paper has been accepted by CVPR 2026!**
* **`[2026/1/27]`:**🔥**We have released the Streamo-Instruct dataset.[[HF](https://huggingface.co/datasets/maifoundations/Streamo-Instruct-465K)].**
* **`[2026/1/22]`:**🔥**We have released our training code.**
* **`[2026/1/6]`:**🔥**We have released our website with more interesting demos [[Web](https://jiaerxia.github.io/Streamo/)].**
* **`[2025/12/24]`:**🔥**We have released our paper [[Arxiv](https://arxiv.org/abs/2512.21334)].**

> **Note:** Due to some restrictions, we are unable to publicly release the model weights at this time. If you have any request, please feel free to contact us.


# Demo🎬

<p align="center">
  <a href="https://youtu.be/lGRdBP-SYeo">
    <img src="https://img.youtube.com/vi/lGRdBP-SYeo/maxresdefault.jpg" alt="Demo Video" width="800">
  </a>
</p>



# Training🚀

## Installation

```bash
pip install -r requirements.txt
```

## Data Format📊

### Raw Data Format

The example raw annotation format in `raw_data.json`:

```json
{
  "video_name": "video1.mp4",
  "video_path": "/path/to/video.mp4",
  "task_type": "QA",
  "source": "custom",
  "question": [
    {"content": "What happens in the video?", "time": "5"}
  ],
  "response": [
    {"content": "A person walks into the room.", "st_time": 5.0, "end_time": 6.0, "time": ""}
  ]
}
```

| Field | Description |
|-------|-------------|
| `question.time` | The second when the question appears (e.g., "5" means `<4s-5s>`) |
| `response.st_time` | Start time of the event (standby begins) |
| `response.end_time` | End time of the event |
| `response.time` | Response time for instant response |

### Training Data Format (Stream Format)

The training data uses a multi-turn conversation format, where each turn corresponds to one video frame (1fps):

```json
{
  "messages": [
    {"role": "system", "content": "System prompt for streaming video assistant"},
    {"role": "user", "content": "Your question\n<0s-1s>\n<stream>"},
    {"role": "assistant", "content": "</Silence>"},
    {"role": "user", "content": "<1s-2s>\n<stream>"},
    {"role": "assistant", "content": "</Standby>"},
    {"role": "user", "content": "<2s-3s>\n<stream>"},
    {"role": "assistant", "content": "</Response> Your answer here"}
  ],
  "videos": ["/path/to/video.mp4"]
}
```

### Data Conversion

Use `scripts/convert_streaming_video.py` to convert raw data to training format:

```bash
# Convert raw_data.json to stream format
python scripts/convert_streaming_video.py to-stream \
    --input raw_data.json \
    --output stream_format.json \
    --video-prefix /path/to/videos \
    --fps 1.0
```

See `dataset/example/` for example files.

### Special Tokens

| Token | Description |
|-------|-------------|
| `</Silence>` | No relevant event or current input is irrelevant |
| `</Standby>` | Event is in progress but not yet completed |
| `</Response>` | Event has completed, start outputting the answer |

### Key Points

- `<stream>` is a placeholder for the current frame, replaced with `<image>` during training
- `<Xs-Ys>` indicates the timestamp interval of the current frame
- Videos are sampled at 1fps, each `<stream>` corresponds to one frame

## Quick Start▶️

```bash
bash train.sh
```


# Acknowledgement

This project is built upon [ms-swift](https://github.com/modelscope/ms-swift). We thank the authors for their excellent work.

# Citation🎓
```
@article{xia2025streaming,
  title={Streaming Video Instruction Tuning},
  author={Xia, Jiaer and Chen, Peixian and Zhang, Mengdan and Sun, Xing and Zhou, Kaiyang},
  journal={arXiv preprint arXiv:2512.21334},
  year={2025}
}
```