---
license: mit
tags:
  - video-interpolation
  - rife
  - frame-interpolation
  - computer-vision
---

# Practical-RIFE Model Mirror

This repository is a **mirror** of the pretrained model weights used by
[Practical-RIFE](https://github.com/hzwer/Practical-RIFE), created to work around
Google Drive's download quota limits when running RIFE-based video interpolation
notebooks (e.g. in Google Colab).

## Credit

All models and the underlying RIFE (Real-Time Intermediate Flow Estimation) method
belong to the original authors. This repo redistributes their released weights
for convenience and reliability only — no modifications have been made to the
model files themselves.

- **Original repository:** https://github.com/hzwer/Practical-RIFE
- **RIFE paper:** Huang, Z., Zhang, T., Heng, W., Shi, B., Zhou, S. (2022).
  *Real-Time Intermediate Flow Estimation for Video Frame Interpolation.* ECCV 2022.
- **Original author:** [hzwer](https://github.com/hzwer)

Please check the [original repository's license](https://github.com/hzwer/Practical-RIFE)
for terms on use and redistribution before using these weights outside personal/research contexts.

## Contents

| File | Description |
|---|---|
| `RIFE_v4.26.zip` | RIFE model, version 4.26 |
| `RIFE_v4.26.heavy.zip` | RIFE model, version 4.26 (heavy variant) |
| `RIFE_v4.25.zip` | RIFE model, version 4.25 |
| `RIFE_v4.25.lite.zip` | RIFE model, version 4.25 (lite variant) |
| `RIFE_v4.22.zip` | RIFE model, version 4.22 |
| `RIFE_v4.22.lite.zip` | RIFE model, version 4.22 (lite variant) |
| `RIFE_v4.9.2.zip` | RIFE model, version 4.9.2 |
| `RIFE_v4.6.zip` | RIFE model, version 4.6 |

## Usage

Download directly with `wget`/`curl`:

```bash
wget https://huggingface.co/Bash2X/RIFE-Models/resolve/main/RIFE_v4.26.zip
```

Or with the `huggingface_hub` Python library:

```python
from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="Bash2X/RIFE-Models",
    filename="RIFE_v4.26.zip"
)
```

## Why this mirror exists

The original Practical-RIFE Colab notebook downloads models from a personal
Google Drive folder, which has a shared daily download quota. When too many
people run the notebook on the same day, downloads fail with a
`FileURLRetrievalError`. Hosting a mirror on Hugging Face avoids this issue
since Hugging Face doesn't impose the same per-file download caps.

## Disclaimer

This is an unofficial mirror maintained for convenience. It is not affiliated
with or endorsed by the original author. If the original author would prefer
this mirror not exist, please open an issue or contact the repo maintainer
and it will be taken down.