RIFE
Real-time intermediate flow estimation for AI video frame interpolation.
About
RIFE (Real-Time Intermediate Flow Estimation for Video Frame Interpolation) is a research project by Zhewei Huang and colleagues from Peking University. Instead of estimating bidirectional optical flows and reversing them, its IFNet network directly estimates intermediate flows coarse-to-fine, reducing artifacts on motion boundaries.
Trained with a privileged distillation scheme, RIFE runs significantly faster than prior flow-based methods and achieves state-of-the-art results on public benchmarks. It became the engine inside many frame interpolation apps, including FlowFrames and ncnn/Vulkan portable builds, and powers smooth slow-motion and framerate upconversion.
What it can do
- Real-time video frame interpolation
- 2x, 4x, 8x, and 16x interpolation
- Direct intermediate flow estimation via IFNet
- Portable ncnn/Vulkan builds for CPU and GPU
Setup
Clone the repo and install requirements, or grab a prebuilt ncnn/Vulkan binary.
Prerequisites
Python and PyTorch for the research code; a GPU recommended.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 103; Trending score 72/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.