High-quality anti-aliasing nodes for video and image processing in ComfyUI workflows. VideoTAADLAA combines TAA with a DLAA-inspired pipeline for sharper, more stable results, while VideoAdaptiveAA focuses on aliasing-prone regions. It uses temporal accumulation, jittered sampling, and CNN-based refinement to improve edge quality.

Important

Disclaimer This node does not use NVIDIA's closed-source SDKs or native DLAA/DLSS binaries. Instead, it is a custom adaptation of Temporal and Spatial Anti-Aliasing techniques commonly found in modern game engines, rebuilt entirely from scratch using PyTorch tensor architecture specifically for ComfyUI video post-processing.

Preview

TAA + DLAA Anti-Aliasing	Adaptive Anti-Aliasing

Models

DLAANet.safetensors is the main 1x DLAA refinement model for reducing aliasing artifacts while preserving edge sharpness. DLAATexture.safetensors is a 1x detail refinement model for restoring fine texture and micro-detail after the DLAA pass.

Note: Both models keep the original image resolution and are not ESRGAN models.

Features

Temporal Anti-Aliasing (TAA): Blends frame history to reduce flicker and sub-pixel jitter.
Neural Reconstruction: DLAA-inspired model for high-quality edge restoration.
Hybrid Pipeline: Combines temporal accumulation with neural reconstruction for stable, high-quality results.
Reduced Ghosting: Motion-aware suppression to minimize trailing artifacts in fast motion.
4-Offset Jitter: Deterministic sub-pixel jitter for sharper and more stable reconstruction.
Edge-Preserving Clarity: Sobel-based detection preserves fine textures while reducing aliasing.
Frequency-Aware Refinement: Uses frequency and perceptual losses for cleaner, sharper results.
Lightweight Model: Optimized for fast inference with minimal VRAM usage.

Comparison

AdaptiveAA Comparison

TAA + DLAA Comparison

🚀 How to use

Add the node to your workflow and choose a preset.

Recommended starting point:

Use Auto for general footage.
Use Balanced for stable video cleanup.
Use Detail for images, hair, fur, wires, and high-detail edges.
Use Smooth for smoother motion and softer temporal blending.

Presets

Presets replace manual parameter tuning and are designed to cover most use cases.

Built-in presets for common use cases:

Auto — balanced default behavior.
Balanced — stable anti-aliasing with moderate sharpening.
Detail — stronger edge and detail reconstruction.
Smooth — smoother temporal blending for softer motion.

Start with Auto, then adjust based on your content.

Installation

Install via ComfyUI Manager (Search for "Anti-Aliasing Pack") or clone this repo into your custom_nodes folder.
Note: The required models will be automatically downloaded when you first run the node.
No additional Python dependencies (requirements) are required.

License

This project is licensed under the MIT License - see the LICENSE file for details.