Practical overview
When it fits
Use it when you can work with a Python repository and pretrained model files and want to generate additional frames from a video, numbered PNG sequence, or image pair.
Who it is for
- Developers who want a Python command-line workflow for video frame interpolation.
- Video editors and researchers exploring frame-rate conversion or slow-motion processing.
- Users who need to generate intermediate frames between two images.
Documented uses
- Increase a clip's frame count with 2× or 4× video interpolation.
- Create a slow-motion result by specifying a target FPS.
- Interpolate a pair of images into a larger set of frames.
Requirements & limits
- The documented setup requires installing Python dependencies and downloading pretrained HD model parameters separately.
- The README's slow-motion FPS option removes audio.
- Processing scale can be adjusted for high-resolution inputs or disordered patterns.
- The README reports 30+ FPS only for 2× 720p interpolation on a 2080 Ti.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Interpolate frames in a video
Generate a higher-frame-rate video from an existing clip using the documented command-line workflow.
Before you start
- Python 3 and the project repository
- Dependencies installed with
pip3 install -r requirements.txt - Pretrained HD model parameters extracted into
train_log/ - A video file, or a numbered PNG image sequence
Steps
Clone the repository and install its requirements.
Download the pretrained HD models, extract them, and place the parameters in
train_log/.Run the command below for 2× interpolation, or use
--exp=2for 4×.python3 inference_video.py --exp=1 --video=video.mp4For an explicit slow-motion target, add an option such as
--fps=60; the README notes that this removes audio.
Expected outcome
The README says the 2× example generates a file named like `video_2X_xxfps.mp4`; the FPS option adds a slow-motion effect.
Guide limitations
- The
--fpsslow-motion option removes audio. - For very high-resolution input such as 4K, the README recommends
--scale=0.5; it suggests trying--scale=2.0if disordered patterns appear. - The repository's stated 30+ FPS example is for 2× 720p interpolation on a 2080 Ti and does not establish performance on other hardware.
Interpolate between two images
Create intermediate PNG frames between a pair of images.
Before you start
- Python 3, the project requirements, and pretrained model parameters in
train_log/ - Two image files
Steps
Install the repository requirements and place the pretrained parameters in
train_log/.Run the command below.
python3 inference_img.py --img img0.png img1.png --exp=4
Expected outcome
The example requests 2^4, or 16×, interpolation results as PNG frames.
Guide limitations
- The README describes generating MP4 or GIF output from the resulting PNGs as a separate FFmpeg step.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- hzwer/ECCV2022-RIFE · Official homepage title · retained in its source language