RIFE: real-time intermediate flow estimation for video frame interpolation

What it supplies: RIFE is the implementation of the ECCV 2022 paper on real-time intermediate flow estimation for video frame interpolation. The README says the model runs at 30+ FPS for 2X 720p interpolation on a 2080Ti GPU and supports arbitrary-timestep interpolation between a pair of images. Useful for: - Double or quadruple the frame rate of a video, or add a slow-motion effect, with inference_video.py. - Interpolate between two still images with inference_img.py and turn the output PNGs into an mp4 or GIF with ffmpeg. - Use the README's list of third-party apps and ports built on RIFE. First action: - Clone the repository and run pip3 install -r requirements.txt. - Download the pretrained HD model and unzip it into train_log/. - Run python3 inference_video.py --exp=1 --video=video.mp4 for 2X interpolation, or --exp=2 for 4X. - For very high resolution such as 4K, add --scale=0.5. Prerequisites and limits: Pretrained models are downloaded separately into train_log/; the README says the HD model is not the one reported in the paper. The README gives no Python or PyTorch versions in the text read; requirements.txt was not read. - When you set --fps for a slow-motion effect, the README says the audio will be removed. - If very high resolution video gives disordered patterns, the README suggests changing --scale rather than guaranteeing a fix. - The README lists third-party software but says the authors are not responsible for it or involved in its development, and some of it is paid. - Quality and speed figures are the authors' own. This README does not state the licence of the code or model in the text read here, so check the repository's licence file before use or redistribution. It says that, according to the open source licence, the authors respect the commercial behaviour of other developers. Video you process carries its own rights. Written from the README text only. Statements are as the README gives them; nothing was installed or run. Source inspected: https://raw.githubusercontent.com/hzwer/ECCV2022-RIFE/5d8adbdd40e12c2c8f91930eff838aebe561c086/README.md ; 11007 bytes ; sha256 05290581ecf28621a7731486fa17ba0e4f02fc689dba0a4ccd2c905f72851498 ; whole README text

{
  "revision": {
    "ref": "commit 5d8adbdd40e12c2c8f91930eff838aebe561c086",
    "pinned": false,
    "source_url": "https://raw.githubusercontent.com/hzwer/ECCV2022-RIFE/5d8adbdd40e12c2c8f91930eff838aebe561c086/README.md"
  },
  "checked_at": "2026-11-07T10:32:40.086Z",
  "usefulness": "- Double or quadruple the frame rate of a video, or add a slow-motion effect, with inference_video.py.\n- Interpolate between two still images with inference_img.py and turn the output PNGs into an mp4 or GIF with ffmpeg.\n- Use the README's list of third-party apps and ports built on RIFE.",
  "source_type": "repository",
  "availability": "available",
  "source_owner": "hzwer (RIFE authors) (declared repository namespace)",
  "known_limitations": [
    "Pretrained models are downloaded separately into train_log/; the README says the HD model is not the one reported in the paper. The README gives no Python or PyTorch versions in the text read; requirements.txt was not read.\n\n- When you set --fps for a slow-motion effect, the README says the audio will be removed.\n- If very high resolution video gives disordered patterns, the README suggests changing --scale rather than guaranteeing a fix.\n- The README lists third-party software but says the authors are not responsible for it or involved in its development, and some of it is paid.\n- Quality and speed figures are the authors' own.\n\nThis README does not state the licence of the code or model in the text read here, so check the repository's licence file before use or redistribution. It says that, according to the open source licence, the authors respect the commercial behaviour of other developers. Video you process carries its own rights.\n\nWritten from the README text only. Statements are as the README gives them; nothing was installed or run. ",
    "Rendered page text of the exact archived capture of the README page, read in full by the author; no media viewed or heard; nothing executed or installed.",
    "Source-documented steps; execution not performed. Internal affiliated review is not independent validation."
  ]
}