RIFE Interpolation

Increase video frame rate with RIFE AI interpolation on Clore.ai

Increase video frame rate with RIFE AI interpolation.

All examples can be run on GPU servers rented through CLORE.AI Marketplace.

Renting on CLORE.AI

1. Visit CLORE.AI Marketplace

2. Filter by GPU type, VRAM, and price

3. Choose On-Demand (fixed rate) or Spot (bid price)

4. Configure your order:

   • Select Docker image

   • Set ports (TCP for SSH, HTTP for web UIs)

   • Add environment variables if needed

   • Enter startup command

5. Select payment: CLORE, BTC, or USDT/USDC

6. Create order and wait for deployment

Access Your Server

• Find connection details in My Orders

• Web interfaces: Use the HTTP port URL

• SSH: ssh -p <port> root@<proxy-address>

What is RIFE?

RIFE (Real-Time Intermediate Flow Estimation) can:

• Increase FPS (24→60, 30→120)

• Create smooth slow motion

• Fix choppy footage

• Real-time processing

Quick Deploy

Docker Image:

pytorch/pytorch:2.5.1-cuda12.4-cudnn9-runtime

Ports:

22/tcp

Command:

pip install torch torchvision && \
git clone https://github.com/megvii-research/ECCV2022-RIFE.git && \
cd ECCV2022-RIFE && \
pip install -r requirements.txt

Installation

Option 1: Python Package

pip install rife-ncnn-vulkan-python

Option 2: From Source

git clone https://github.com/megvii-research/ECCV2022-RIFE.git
cd ECCV2022-RIFE
pip install -r requirements.txt

Basic Usage

Double Frame Rate

python inference_video.py --exp=1 --video=input.mp4

# Output: 2x FPS

4x Frame Rate

python inference_video.py --exp=2 --video=input.mp4

# Output: 4x FPS

8x Frame Rate

python inference_video.py --exp=3 --video=input.mp4

# Output: 8x FPS

Python API

Load Model

import torch
from model.RIFE import Model

device = torch.device("cuda")
model = Model()
model.load_model('./train_log', -1)
model.eval()
model.device()

Interpolate Single Frame

import cv2
import numpy as np
import torch

def interpolate_frames(frame1, frame2, model, num_frames=1):
    """Interpolate frames between two images"""
    # Prepare tensors
    img0 = torch.from_numpy(frame1).permute(2, 0, 1).float() / 255.0
    img1 = torch.from_numpy(frame2).permute(2, 0, 1).float() / 255.0

    img0 = img0.unsqueeze(0).cuda()
    img1 = img1.unsqueeze(0).cuda()

    # Interpolate
    with torch.no_grad():
        middle = model.inference(img0, img1)

    # Convert back
    middle = (middle[0] * 255).byte().cpu().numpy().transpose(1, 2, 0)
    return middle

# Load frames
frame1 = cv2.imread('frame1.png')
frame2 = cv2.imread('frame2.png')

# Get interpolated frame
middle_frame = interpolate_frames(frame1, frame2, model)
cv2.imwrite('interpolated.png', middle_frame)

Process Video

import cv2
import torch
from model.RIFE import Model

model = Model()
model.load_model('./train_log', -1)
model.eval()
model.device()

def process_video(input_path, output_path, multiplier=2):
    cap = cv2.VideoCapture(input_path)
    fps = cap.get(cv2.CAP_PROP_FPS)
    width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

    out = cv2.VideoWriter(
        output_path,
        cv2.VideoWriter_fourcc(*'mp4v'),
        fps * multiplier,
        (width, height)
    )

    ret, prev_frame = cap.read()
    if not ret:
        return

    out.write(prev_frame)

    while True:
        ret, curr_frame = cap.read()
        if not ret:
            break

        # Prepare tensors
        img0 = torch.from_numpy(prev_frame).permute(2, 0, 1).float() / 255.0
        img1 = torch.from_numpy(curr_frame).permute(2, 0, 1).float() / 255.0

        img0 = img0.unsqueeze(0).cuda()
        img1 = img1.unsqueeze(0).cuda()

        # Generate intermediate frames
        for i in range(multiplier - 1):
            t = (i + 1) / multiplier
            with torch.no_grad():
                middle = model.inference(img0, img1, timestep=t)
            middle = (middle[0] * 255).byte().cpu().numpy().transpose(1, 2, 0)
            out.write(middle)

        out.write(curr_frame)
        prev_frame = curr_frame

    cap.release()
    out.release()

process_video('input.mp4', 'output_60fps.mp4', multiplier=2)

Using rife-ncnn-vulkan

Faster NCNN implementation:

from rife_ncnn_vulkan import Rife

rife = Rife(gpu_id=0)

# Interpolate
frame1 = Image.open('frame1.png')
frame2 = Image.open('frame2.png')
middle = rife.process(frame1, frame2)
middle.save('interpolated.png')

Video Processing

from rife_ncnn_vulkan import Rife
import cv2
from PIL import Image

rife = Rife(gpu_id=0, num_threads=4)

def interpolate_video(input_path, output_path, factor=2):
    cap = cv2.VideoCapture(input_path)
    fps = cap.get(cv2.CAP_PROP_FPS)
    width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

    out = cv2.VideoWriter(
        output_path,
        cv2.VideoWriter_fourcc(*'mp4v'),
        fps * factor,
        (width, height)
    )

    ret, prev = cap.read()
    out.write(prev)

    while True:
        ret, curr = cap.read()
        if not ret:
            break

        # Convert to PIL
        prev_pil = Image.fromarray(cv2.cvtColor(prev, cv2.COLOR_BGR2RGB))
        curr_pil = Image.fromarray(cv2.cvtColor(curr, cv2.COLOR_BGR2RGB))

        # Interpolate
        for i in range(factor - 1):
            t = (i + 1) / factor
            mid = rife.process(prev_pil, curr_pil, timestep=t)
            mid_cv = cv2.cvtColor(np.array(mid), cv2.COLOR_RGB2BGR)
            out.write(mid_cv)

        out.write(curr)
        prev = curr

    cap.release()
    out.release()

Slow Motion

Create smooth slow motion:

# Original: 30 fps, 10 seconds = 300 frames

# 8x interpolation: 2400 frames

# Play at 30 fps: 80 seconds (8x slower)

python inference_video.py --exp=3 --video=input.mp4

# This creates 8x frames, play at original FPS for slow motion

Slow Motion Script

def create_slow_motion(input_path, output_path, slowdown_factor=4):
    """Create slow motion video"""
    cap = cv2.VideoCapture(input_path)
    original_fps = cap.get(cv2.CAP_PROP_FPS)

    # Interpolate to get more frames
    exp = int(np.log2(slowdown_factor))
    interpolate_video(input_path, 'temp_interpolated.mp4', factor=slowdown_factor)

    # Re-encode at original FPS
    cap2 = cv2.VideoCapture('temp_interpolated.mp4')
    width = int(cap2.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(cap2.get(cv2.CAP_PROP_FRAME_HEIGHT))

    out = cv2.VideoWriter(
        output_path,
        cv2.VideoWriter_fourcc(*'mp4v'),
        original_fps,  # Keep original FPS
        (width, height)
    )

    while True:
        ret, frame = cap2.read()
        if not ret:
            break
        out.write(frame)

    cap.release()
    cap2.release()
    out.release()

Batch Processing

import os
from concurrent.futures import ThreadPoolExecutor

def process_single(input_path, output_dir, factor=2):
    filename = os.path.basename(input_path)
    output_path = os.path.join(output_dir, f"interpolated_{filename}")
    interpolate_video(input_path, output_path, factor)
    return output_path

input_dir = './videos'
output_dir = './interpolated'
os.makedirs(output_dir, exist_ok=True)

videos = [os.path.join(input_dir, f) for f in os.listdir(input_dir)
          if f.endswith(('.mp4', '.mkv', '.avi'))]

for video in videos:
    result = process_single(video, output_dir, factor=2)
    print(f"Completed: {result}")

Quality Settings

Model Versions

Model	Quality	Speed
RIFE v4.6	Best	Slow
RIFE v4.0	Great	Medium
RIFE-NCNN	Good	Fastest

UHD Mode

For 4K+ videos:

python inference_video.py --exp=1 --video=input.mp4 --UHD

Memory Optimization

For Limited VRAM

# Process in tiles
from model.RIFE import Model

model = Model()
model.load_model('./train_log', -1)
model.eval()

# Set tile size for large frames

# model.inference handles tiling internally

Reduce Memory

# Use NCNN version (more memory efficient)
pip install rife-ncnn-vulkan-python

Performance

Resolution	GPU	2x Interp FPS
1080p	RTX 3090	~60 fps
1080p	RTX 4090	~100 fps
4K	RTX 3090	~15 fps
4K	RTX 4090	~30 fps

Troubleshooting

Artifacts/Ghosting

• Use scene detection to skip cuts

• Reduce interpolation factor

• Check for fast motion

Out of Memory

• Use NCNN version

• Process at lower resolution, upscale after

• Reduce batch size

Slow Processing

• Use NCNN-Vulkan version

• Enable GPU acceleration

• Use smaller model

Scene Detection

Skip interpolation across scene cuts:

from scenedetect import detect, ContentDetector

scenes = detect('input.mp4', ContentDetector())

# Don't interpolate between scenes
for scene in scenes:
    print(f"Scene: {scene[0].get_frames()} - {scene[1].get_frames()}")

Cost Estimate

Typical CLORE.AI marketplace rates (as of 2024):

GPU	Hourly Rate	Daily Rate	4-Hour Session
RTX 3060	~$0.03	~$0.70	~$0.12
RTX 3090	~$0.06	~$1.50	~$0.25
RTX 4090	~$0.10	~$2.30	~$0.40
A100 40GB	~$0.17	~$4.00	~$0.70
A100 80GB	~$0.25	~$6.00	~$1.00

Prices vary by provider and demand. Check CLORE.AI Marketplace for current rates.

Save money:

• Use Spot market for flexible workloads (often 30-50% cheaper)

• Pay with CLORE tokens

• Compare prices across different providers

Next Steps

• FFmpeg NVENC - Encode output

• Real-ESRGAN - Upscale video

• AI Video Generation - Generate videos