yuvraj108c/ComfyUI-Depth-Anything-Tensorrt

Leverages NVIDIA TensorRT engine compilation with FP16 precision to accelerate depth estimation inference, supporting multiple model variants (v1/v2/v3 and Distill Any Depth) across different scales. Integrates directly into ComfyUI workflows via custom nodes with automated engine building, offering both basic and advanced modes with flexible depth map visualization using configurable colormaps and adjustment parameters. Optimized for NVIDIA GPUs (RTX4090, RTX5090, H100) with performance reaching 200+ FPS on high-end hardware through native TensorRT deployment.

127 stars.