Civitai #145349049: High-Contrast Penthouse Skyline Silhouette - Krea 2 Turbo & Qwen3-VL

Full Resolution

Generation Specs RTX 4070

Sampler er_sde

Scheduler simple

Steps 9

CFG Scale 1

Seed 8451825544000725

Target VRAM 12GB

VRAM 10GB - 12GB

GPU Compatibility

Optimal Performance

Runs entirely in VRAM with no memory swapping.

Krea 2 Turbo (Official Comfy-Org) Civitai #145349049 Min 10GB VRAM 944×1664 RTX 4070 Verified

High-Contrast Penthouse Skyline Silhouette - Krea 2 Turbo & Qwen3-VL

High-contrast golden hour penthouse silhouette featuring an East Asian woman in a black silk evening gown against panoramic skyline windows rendered on Krea 2 Turbo.

Blueprint Summary RTX 4070 Verified

Reproducible ComfyUI workflow for Civitai #145349049: High-Contrast Penthouse Skyline Silhouette - Krea 2 Turbo & Qwen3-VL using Krea 2 Turbo (Official Comfy-Org) at 944×1664 resolution. Requires minimum 10GB VRAM with sampler er_sde and scheduler simple (9 steps). Includes 1-click terminal model sync and canvas JSON graph.

Pipeline Flow 9 Nodes

Open in Resolver

01 Load Models

UNETLoader

3 loaders (DiT, CLIP, VAE)

02 Conditioning

CLIPTextEncode

2 prompt encodings

03 KSampler

KSampler

DiT latent denoising

04 Decode & Save

VAEDecode

Latent to pixel space

Execution DAG Topology Interactive Visualizer

Drag to pan · Scroll to zoom · Hover wires

Model & Asset Setup Setup Script

Run in your ComfyUI root:

curl -fsSL https://decomfy.com/api/scripts/civitai-penthouse-skyline-silhouette-krea2.sh | bash

Positive Prompt

Cinematic high-contrast editorial fashion photograph, vertical 9:16 framing. Camera low behind, 35mm lens at f/2.8 with dramatic shallow depth of field. A graceful East Asian woman with long glossy auburn-brown hair standing elegantly before an expansive floor-to-ceiling penthouse window. She wears a floor-length tailored black silk crepe evening gown with sleek long sleeves and architectural minimalist drape. Her arms gently rest near the glass as she gazes out at the glowing morning skyline. Intense natural rim backlight from the towering sunlit window creating a striking high-fashion silhouette, with luminous golden rim lighting tracing her shoulders, arms, and hair contours. Fine tactile 35mm film grain, deep obsidian contrast, rich velvety shadow tones with subtle fabric sheen. Clean minimalist penthouse interior, polished dark hardwood floor reflections. Anatomically flawless, five fingers on each hand, serene and contemplative atmosphere. No text, watermark, logo, or distortion.

Negative Prompt

unclothed, disrobed, revealing clothes, lowres, blurry, bad anatomy, deformed hands, extra fingers, low quality, artifacts, distorted proportions, text, watermark, signature

Prompt Customization: This workflow is pre-wired and calibrated. Swap this prompt with your own character, subject, or style without breaking the pipeline.

Required Models 3 Models

Required Storage: 17.7 GB (3 models · DiT/Base: 12.6 GB · Text Encoder: 4.9 GB · VAE: 253 MB)

models/diffusion_models/ 12.56 GB HuggingFace

krea2_turbo_int8_convrot.safetensors Official Krea 2 Turbo INT8 ConvRot Diffusion Transformer core

Get

models/text_encoders/ 4.88 GB HuggingFace

qwen3vl_4b_fp8_scaled.safetensors Official Qwen3-VL 4B FP8 Scaled Text Encoder (CLIPLoader type: krea2)

Get

models/vae/ 253 MB HuggingFace

qwen_image_vae.safetensors Official 16-channel Video & Image VAE

Get

Field Notes RTX 4070 Benchmark

Benchmark: Rendered on NVIDIA GeForce RTX 4070 (12GB VRAM, CUDA 12.4)

1. ARCHITECTURAL REASONING & COMPOSITION GEOMETRY: Deconstructs a viral high-contrast penthouse silhouette portrait into a pristine SFW luxury lookbook editorial. Krea 2 Turbo architecture paired with the Qwen3-VL 4B text encoder handles complex backlight rim lighting with extreme precision. The sharp contrast between the dark interior silhouette and the sun-drenched skyscraper horizon highlights the drape of the floor-length black silk crepe gown without sacrificing garment texture or fine hand anatomy on the glass. 2. CABLE ROUTING & SIGNAL TOPOLOGY: Streamlined single-pass topology connecting UNETLoader (krea2_turbo_int8_convrot.safetensors, fp8_e4m3fn) and CLIPLoader (qwen3vl_4b_fp8_scaled.safetensors, type: krea2). A single deterministic Positive conditioning node feeds directly into KSampler along with Universal Clean Negatives. Sampling is executed via er_sde with simple scheduler across 9 steps at CFG 1.0, decoding cleanly through the 16-channel qwen_image_vae.safetensors into a 944x1664 latent. 3. HARDWARE BENCHMARKS & CONSUMER GPU FOOTPRINT: Executed locally on host NVIDIA GeForce RTX 4070 (12GB VRAM, CUDA 12.4). Peak VRAM consumption reached 11.84 GB during the 9-step er_sde sampling run with zero paging into host system RAM. Total generation latency clocked at 89.4 seconds (~5.8s per step + VAE decode).

Frequently Asked Questions FAQ

What GPU and VRAM are required to run Civitai #145349049: High-Contrast Penthouse Skyline Silhouette - Krea 2 Turbo & Qwen3-VL?

This workflow requires a minimum of 10GB VRAM (recommended 12GB VRAM). Tested and verified on NVIDIA GeForce RTX 4070 (12GB VRAM) at 944x1664 resolution.

How do I resolve missing custom nodes for this workflow?

You can drop the workflow JSON into our client-side Missing Node Auto-Resolver at https://decomfy.com/resolve/ to detect missing nodes and generate install commands, or run the 1-click terminal setup script provided below.

Which VAE and conditioning text encoders are required for Krea 2 DiT?

Krea 2 DiT single-stream models strictly require the 16-channel Wan2.1 VAE and Qwen3-VL text encoder. Using standard Flux or SDXL VAE causes inverted checkerboard artifacts. Also pair with RescaleCFG for high-contrast portrait clarity.