An advanced ComfyUI workflow that combines intelligent batch upscaling with AI frame interpolation to transform low-resolution videos into high-resolution, buttery-smooth masterpieces without memory errors.
This workflow represents the ultimate post-processing solution for AI-generated videos. It not only increases resolution through advanced upscaling but also dramatically improves motion smoothness by interpolating frames to 32FPS or higher. All while maintaining the batch processing that prevents Out of Memory errors on long videos.
📖 Description
Why choose between resolution and smoothness when you can have both? This workflow is the natural evolution of batch upscaling, adding professional-grade frame interpolation to create truly cinematic results. It takes your AI-generated videos (from Wan2.2, AnimateDiff, etc.) and performs a two-stage enhancement: first doubling the resolution with RealESRGAN, then using RIFE AI interpolation to multiply the frame rate for incredibly fluid motion.
Perfect for transforming jerky 16FPS AI animations into smooth 32FPS+ cinematic sequences that rival professional animation quality.
✨ Features & Highlights
🌟 Dual Enhancement: Combines resolution upscaling AND frame interpolation in one seamless workflow
🚫 OOM-Proof Design: Inherits the batch processing system that prevents memory errors on videos of any length
🎬 Professional Smoothness: Uses RIFE VFI with
rife49.pthmodel for state-of-the-art frame interpolation⚡ Flexible Output: Easily adjustable output frame rate (default 32FPS, configurable up to 60FPS+)
🎵 Perfect Audio Sync: Maintains original audio quality and synchronization throughout the process
🔧 Model Swappable: Compatible with any upscaling model (2x, 4x) and interpolation model
📊 Smart Batching: Processes 16 frames per batch by default, optimized for most GPUs
🛠️ Technical Details
Workflow: ComfyUI (JSON included)
Required Custom Nodes:
ComfyUI-VideoHelperSuite(for batch video processing)comfyui-frame-interpolation(for RIFE VFI)
Core Enhancement Pipeline:
text
Input Video → Batch Loading → 2x Upscaling → Frame Interpolation → Smooth Video OutputDefault Models & Settings:
Upscale Model:
RealESRGAN_x2plus.pth(2x resolution increase)Interpolation Model:
rife49.pth(high-quality frame interpolation)Multiplier: 2x (doubles frame rate)
Output Frame Rate: 32 FPS (from typical 16 FPS input)
Batch Size: 16 frames per processing chunk
🚀 Usage Instructions
Load the Workflow: Import the provided JSON into ComfyUI.
Install Required Nodes: Ensure you have Video Helper Suite and Frame Interpolation installed.
Select Input Video: In the
VHS_LoadVideonode, upload your source video file.Adjust Settings (Optional):
Change
frames_per_batchinVHS_BatchManagerfor your GPU (smaller for weaker cards)Modify the multiplier in
RIFE VFInode for different smoothness levels (2x, 3x, 4x)Adjust output
frame_rateinVHS_VideoCombineto match your interpolation
Run the Workflow: Execute and watch as your video is enhanced in two stages.
💡 Workflow Breakdown
The workflow follows a sophisticated but logical chain:
Video Loading & Batching:
VHS_LoadVideoloads the source video and extracts audioVHS_BatchManagerdivides the video into manageable 16-frame chunks
Resolution Enhancement:
UpscaleModelLoaderloads the RealESRGAN 2x modelImageUpscaleWithModelupscales each frame, doubling the resolution
Motion Smoothing:
RIFE VFItakes the upscaled frames and generates intermediate framesUses AI to create smooth transitions between original frames
Applies 2x multiplier (16FPS → 32FPS)
Final Assembly:
VHS_VideoCombinereassembles all processed framesSyncs with original audio at the new frame rate (32FPS)
⚙️ Recommended Settings
For Maximum Quality:
Use
RealESRGAN_x4plus.pthfor 4x upscalingSet RIFE multiplier to 2x for 32FPS output
For Weaker GPUs:
Reduce batch size to 8-10 frames
Use 2x upscaling instead of 4x
Consider
rife46.pthfor faster interpolation
For Ultra-Smooth Results:
Set RIFE multiplier to 3x or 4x (48FPS/64FPS)
Ensure output frame rate matches in VideoCombine node
Popular Model Combinations:
Balanced:
RealESRGAN_x2plus.pth+rife49.pth(2x)Quality Focused:
4x-UltraSharp.pth+rife49.pth(2x)Speed Focused:
RealESRGAN_x2plus.pth+rife46.pth(3x)
🎯 Ideal Use Cases
AI Video Enhancement: Transform Wan2.2/AnimateDiff outputs from jerky to cinematic
Animation Smoothing: Perfect for cartoon, anime, or 3D animation frame rate improvement
Game Recording Enhancement: Smooth out low-FPS game recordings
Historical Footage Restoration: Upscale and smooth old low-frame-rate videos
Social Media Content: Create buttery-smooth videos for TikTok, YouTube, etc.
⚠️ Limitations & Notes
Processing Time: The dual enhancement process takes significant time (quality takes patience)
Artifact Management: High interpolation can sometimes create smoothing artifacts in fast motion
Audio Duration: The original audio length remains unchanged - video may loop if frames exceed audio
Model Compatibility: Ensure your interpolation model supports desired multiplier factors
Storage Requirements: Temporary files require substantial disk space for long videos
Pro Tip: For best results, start with a high-quality source video. The better the input, the more spectacular the output!
🆚 Comparison with Basic Upscaler
While the basic upscaler workflow provides resolution improvement, this enhanced version adds:
Motion smoothing through AI interpolation
Higher perceived quality from increased frame rate
More cinematic results suitable for professional use
Flexible output options (adjustable frame rates)
Tags: ComfyUI, Workflow, Video Upscaling, Frame Interpolation, RIFE, Smooth Video, OOM Fix, Video Enhancement, AI Video, Post-Processing
Credit: uscha (Based on the file path in the workflow)
This workflow represents the cutting edge of AI video post-processing, delivering professional-grade results that elevate your AI-generated content to truly cinematic quality!
Description
Enhanced output quality and increased stability have been achieved.