Frame Interpolation & Video Upscaling directly in After Effects.
Remove duplicates, make video smoother, and make clips larger using your graphics card.
AniSmooth is a free local After Effects extension built to help you remove duplicate frames, make videos smoother with RIFE, and upscale video clips locally on your computer with a Python backend.
Unlike other large and heavy extensions, AniSmooth is built with a small, clean, and highly compact codebase. The code is split into simple, organized files to ensure the extension loads instantly and runs smoothly alongside other tools without slowing down After Effects.
Note
Compatibility: Supports After Effects CC 2018 through CC 2026+ (v15.0+). Windows primary support.
- Threshold Slider - Adjust how sensitive the check is for duplicate frames.
- Motion Tracking - Tells the difference between slow motion and frozen frames.
- Frozen Character Detection - Finds frames where the background moves but characters are completely frozen.
- RIFE Models - Supports RIFE 4.25 models (running on CUDA or TensorRT).
- Custom Multipliers - Choose speed multipliers (2x, 4x, 6x, 8x, 10x) or write a custom number up to 64x.
- Upscaling Multipliers - Make clips 2x or 4x larger.
- Model Support - ShuffleCUGAN models built for anime and detailed clips.
- Usage Monitor - Shows RAM usage, GPU usage, and GPU temperature.
- Setup Wizard - Installer checker that helps you download and set up Python, FFmpeg, and model files.
- GPU Diagnostics - Shows VRAM usage and your graphics card model.
- Output Preferences - Set where files save, customize names, prevent overwriting files, and automatically import completed clips into your composition.
- Interface Toggles - Hide tabs you do not use and choose which models appear in the dropdown list.
- Config Presets - Save, import, and export your settings to share them or keep backups.
AniSmooth supports local hardware-accelerated models for both frame interpolation and video upscaling.
Note
Beta Stage: We plan to add more models in the future. Right now, because the extension is in beta, we are keeping only these selected models - which are currently considered the best ones available - until we make sure the extension is 100% stable.
| Model Key | Model Name | Parameters | VRAM Required | Engine / Acceleration | Description & Use Case |
|---|---|---|---|---|---|
rife4.25 |
RIFE 4.25 Cuda | 1.3M | ~2GB | PyTorch CUDA | Fast, lightweight model for quick renders and older graphics cards. |
rife4.25-heavy |
RIFE 4.25 HEAVY Cuda | 5.1M | ~6GB | PyTorch CUDA | Larger model that gives the best motion results but needs a stronger graphics card. |
rife4.25-tensorrt |
RIFE 4.25 TensorRT | 1.3M | ~2GB | NVIDIA TensorRT | Optimized version that runs up to 1.8x faster on compatible NVIDIA cards. |
rife4.25-heavy-tensorrt |
RIFE 4.25 HEAVY TensorRT | 5.1M | ~6GB | NVIDIA TensorRT | The high-quality model optimized to run much faster on NVIDIA cards. |
| Model Key | Model Name | Parameters | VRAM Required | Engine / Acceleration | Description & Use Case |
|---|---|---|---|---|---|
adore |
Adore Cuda | 4M | ~3GB | PyTorch CUDA | Keeps lines sharp and retains details when upscaling. |
fallin_soft |
Fallin Soft Cuda | 3.9M | ~4GB | PyTorch CUDA | Built for anime, making colors smooth, backgrounds clean, and lines look sharp. |
- Pick your After Effects version folder (AE2018 / AE2020 / AE2022).
- Download ZXP Installer (Windows & macOS).
- Drag
AniSmooth_AE2020.zxp(or corresponding version) onto the ZXP Installer window. - Restart After Effects, go to
Window > Extensions > AniSmooth.
- Open your version folder and run
AniSmoothSetup_AE2020.exe. - Follow the setup wizard - it handles file placement and registry keys automatically.
- Copy the
AniSmoothfolder from your desired version folder to:
- Windows:
C:\Program Files (x86)\Common Files\Adobe\CEP\extensions\
- Enable PlayerDebugMode: double-click
Add-Keys.regor runAdd-Keys.batas admin. - Restart After Effects.
cd tools && npm install
cd .. && npm run build:allThis extension runs local Python and AI models on your system. Make sure you meet the VRAM requirements (2GB+ for basic, 6GB+ for heavy models) and have an NVIDIA GPU for CUDA acceleration.






