Realistic SDXL CompfyUI Workflow vs Leonardo AI

I have made a quick comparison between Leonardo AI (Finetuned 1.5 realistics model) and SDXL workflow (SDXL 1.0 based, SDXL 1.0 refiner, SDXL 1.0 VAE) with Leonardo by inputing the same prompt for both (it is actually sample prompt from Leonardo AI)

Prompt: highly detailed digital art masterpiece, artwork design majestic face of a beautiful Russian girl, steampunk, sharp focus, real and colorful shades, highly intricate and detailed, photorealistic masterpiece, professional photography, cinematic lighting, hd 8 k

Negative : out of frame, clipped, cropped image, bad proportion, double, duplicated, deformed, disfigured, plastic, blurred, bad anatomy, ugly, grainy, low resolution, badly drawn face arms hands legs fingers, mutation, extra limb, missing limb, floating limbs, detached limbs, long neck body, out of focus, text, writing, letters, watermark

Leanardo result on https://app.leonardo.ai/

Master image

Prompting and model config

SDXL Workflow result (On premised hosting on NamiQ center with Xeon E5 single CPU, A4000 GPU workstation)

Master image

Workflow configuration

It's challenge to judge which is better since the output is slightly same but different coloring style (like the story of Samsung and Sony display's color).

I am voting for ComfyUI SDXL 1.0 as it is free version but providing the comparative output and controllable workflow (thanks for ComfyUI) while Leonardo AI provides the easy to use commercialized platform.

Professional user with technical installation experience might like ComfyUI SDXL an alternative to host and make SDXL accessible and more controllable internally.