ComfyUI Upscaler TensorRT

This project provides a Tensorrt implementation for fast image upscaling inside ComfyUI (3-4x faster)

This project is licensed under CC BY-NC-SA, everyone is FREE to access, use, modify and redistribute with the same license.

For commercial purposes, please contact me directly at yuvraj108c@gmail.com

If you like the project, please give me a star! ⭐

---

⏱️ Performance

Note: The following results were benchmarked on FP16 engines inside ComfyUI, using 100 frames

Device	Model	Input Resolution (WxH)	Output Resolution (WxH)	FPS
L40s	RealESRGAN_x4	512 x 512	2048 x 2048	5
L40s	RealESRGAN_x4	960 x 540	3840 x 2160	2
L40s	RealESRGAN_x4	1280 x 1280	5120 x 5120	0.7

🚀 Installation

Navigate to the ComfyUI /custom_nodes directory

git clone https://github.com/yuvraj108c/ComfyUI-Upscaler-Tensorrt.git
cd ./ComfyUI-Upscaler-Tensorrt
pip install -r requirements.txt

🛠️ Building Tensorrt Engine

1. Download one of the available onnx models. These onnx models support dynamic image resolutions from 256x256 to 1280x1280 px (e.g 960x540, 512x512, 1280x720 etc). Here are the original models:

   • 4x-AnimeSharp
   • 4x-UltraSharp
   • 4x-WTP-UDS-Esrgan
   • 4x_NMKD-Siax_200k
   • 4x_RealisticRescaler_100000_G
   • 4x_foolhardy_Remacri
   • RealESRGAN_x4

2. Run python export_trt.py and set onnx/engine paths accordingly

3. Place the exported engine inside ComfyUI /models/tensorrt/upscaler directory

☀️ Usage

• Insert node by Right Click -> tensorrt -> Upscaler Tensorrt
• Choose the appropriate engine from the dropdown

⚠️ Known issues

• Only models with ESRGAN architecture are currently working
• High ram usage when exporting .pth to .onnx

🤖 Environment tested

• Ubuntu 22.04 LTS, Cuda 12.3, Tensorrt 10.0.1, Python 3.10, L40s GPU
• Windows 11

👏 Credits

• NVIDIA/Stable-Diffusion-WebUI-TensorRT
• comfyanonymous/ComfyUI

License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)