Simple SDXL ControlNet Upscaler Workflow for ComfyUI

A high-quality ComfyUI workflow using RealVisXL V5.0, enhanced with ControlNet Depth preprocessing and 4x UltraSharp upscaling.

🎯 Main Features:

- 🧠 Model: realvisxlV50_v50Bakedvae.safetensors

- 📐 ControlNet: diffusion_pytorch_model_promax.safetensors using DepthAnythingPreprocessor

- 🔍 Upscaler: 4x-UltraSharp.pth

- 🔁 Sampler: dpmpp_2m_sde with Karras scheduler

- ⚙️ Denoising strength: 0.7

- 🖼️ Output resolution: 512x512 (then upscaled)

🧩 This workflow uses:

- comfyui_controlnet_aux for preprocessing

- ComfyUI-Easy-Use for user-friendly nodes

- ComfyUI-Crystools for extra utilities

- rgthree's ComfyUI Nodes for advanced prompt and model control

Example prompt: "dreamy pastel beach, minimalist aesthetic, california coastal vibes, soft lighting, serene sunset, pink and orange sky, elegant composition, light reflections, ocean background, aesthetic photography, soft sand texture, vintage summer tones, Fujifilm XT3, RAW photo, high quality, 8k uhd, dslr, film grain"

Negative prompt: "(deformed iris, deformed pupils, semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, anime), text, watermark, logo, cropped, out of frame, low quality, jpeg artifacts, blurry, poorly drawn hands, poorly drawn face, mutation, bad anatomy, extra limbs, fused fingers, UnrealisticDream"