Files
SOLRICKS-comfyui-solricks/README.md
T
MSXYZ 95d95b9faa docs: add comparison video
docs: add comparison video
2026-04-22 18:10:31 +03:00

89 lines
3.9 KiB
Markdown

# ComfyUI Video Anti-Aliasing Pack
<p align="center">
<img src="https://img.shields.io/badge/ComfyUI--Manager-Verified-green?style=flat-square&logo=github" alt="Manager">
<img src="https://img.shields.io/github/v/release/MSXYZ-GenAI/comfyui-msxyz?style=flat-square&color=orange" alt="Release">
<img src="https://img.shields.io/badge/python-3.10+-blue?style=flat-square&logo=python" alt="Python">
<img src="https://img.shields.io/github/stars/MSXYZ-GenAI/comfyui-msxyz?style=flat-square&color=gold" alt="Stars">
</p>
Advanced anti-aliasing nodes for videos and images. VideoTAADLAA combines TAA with a DLAA-inspired pipeline for sharp, stable results, while VideoAdaptiveAA targets specific aliasing-prone regions. The system utilizes temporal accumulation, jittered sampling, and CNN-based refinement for superior 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 |
| :---: | :---: |
| ![Video TAA + DLAA Node](https://raw.githubusercontent.com/MSXYZ-GenAI/comfyui-msxyz/refs/heads/main/video_taadlaa.png) | ![Video Adaptive AA Node](https://github.com/MSXYZ-GenAI/comfyui-msxyz/blob/main/video_adaptive_aa.png) |
---
## Features
- **Temporal Anti-Aliasing (TAA):** Blends frame history with motion data to eliminate flicker and sub-pixel jitter.
- **Neural Reconstruction (MSE 0.000023):** DLAA-inspired model for near-lossless, high-fidelity edge restoration.
- **Ghosting-Free Tech:** Motion-aware suppression logic to minimize trailing artifacts in high-speed sequences.
- **4-Offset Jitter:** Deterministic sub-pixel sampling for sharper and more consistent reconstruction stability.
- **Edge-Preserving Clarity:** Sobel-based detection preserves fine textures while eliminating aliasing artifacts.
- **Frequency-Aware Refinement:** Frequency and Perceptual loss optimization for crystal-clear, blur-free results.
---
## Comparison
<p align="center">
<strong>AdaptiveAA Comparison</strong>
</p>
<table align="center">
<tr>
<td align="center"><strong></strong></td>
<td align="center"><strong></strong></td>
</tr>
<tr>
<td><img src="compare_1.png" width="450"></td>
<td><img src="compare_2.png" width="450"></td>
</tr>
</table>
<br>
<p align="center">
<strong>TAA + DLAA Comparison</strong>
<video src="https://github.com/user-attachments/assets/ec5d576e-f328-430d-8051-fc65800af4e1" width="500" controls autoplay muted loop>
</video>
</p>
---
## 🚀 How to use
> - Test the effect by increasing `dlaa_strength` to **0.8** or **1.0**.
**Parameters**
- **taa_strength:** Controls the impact of past frames. (Default: 0.6 | Fast Motion: 0.4 | Static: 0.9)
- **taa_alpha:** Determines how quickly history updates. (Default: 0.1 | Aggressive AA: 0.05 | Ghost-free: 0.4)
- **motion_sensitivity:** Threshold to disable TAA in moving areas. (Default: 0.08 | Flicker-free: 0.0 | High-speed: 0.2)
- **jitter_scale:** Amount of sub-pixel shifting for reconstruction. (Default: 0.2 | Off: 0.0 | Maximum: 0.5)
- **dlaa_strength:** Power of the neural reconstruction model. (Default: 0.8 | Subtle: 0.4 | Aggressive: 1.0)
- **edge_threshold:** Sensitivity of edge detection. (Default: 0.15 | Hard edges: 0.3 | Fine details: 0.05)
- **blur_radius:** Radius of pre-AA edge softening. (Default: 1 | Sharp: 0 | Softer: 2)
- **reset_history:** Clears temporal memory. (Default: False | Enable on scene cuts: True)
---
## Installation
Recommended: Install via ComfyUI-Manager.
Manual installation:
```bash
cd ComfyUI/custom_nodes/
git clone https://github.com/MSXYZ-GenAI/comfyui-msxyz.git
```
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.