Krea 2 style transfer ComfyUI workflow | Reference to Scene Stylization

Turn any reference style into new stunning visual scenes.

Who it's for: creators who want this pipeline in ComfyUI without assembling nodes from scratch. Not for: one-click results with zero tuning - you still choose inputs, prompts, and settings.

Open preloaded workflow on RunComfy

Open preloaded workflow on RunComfy (browser)

Why RunComfy first
- Fewer missing-node surprises - run the graph in a managed environment before you mirror it locally.
- Quick GPU tryout - useful if your local VRAM or install time is the bottleneck.
- Matches the published JSON - the zip follows the same runnable workflow you can open on RunComfy.

When downloading for local ComfyUI makes sense - you want full control over models on disk, batch scripting, or offline runs.

How to use (local ComfyUI)
1. Load inputs (images/video/audio) in the marked loader nodes.
2. Set prompts, resolution, and seeds; start with a short test run.
3. Export from the Save / Write nodes shown in the graph.

Expectations - First run may pull large weights; cloud runs may require a free RunComfy account.

---

Overview

With this style transfer workflow, you can map the visual language of any reference image onto an entirely new composition. It lets you experiment with poster, clay, oil, watercolor, anime, or photographic looks without copying original layouts. Designed for artists and designers, it ensures style consistency across unique subjects. Enhanced by Qwen3-VL encoding and Krea 2 Turbo, it produces accurate, stable, and high-fidelity aesthetic results. Ideal for creatives who need fast, precise, and versatile style recreation.

Important nodes:

Key nodes in ComfyUI Krea 2 style transfer ComfyUI workflow

RFInversion (#603)

Role: learns a compact representation of the reference image’s style and returns a lightly adapted model plus a style latent. Increase its influence if the style feels too subtle, or reduce it if the style overwhelms subject identity. Backed by the Untwisting RoPE implementation for ComfyUI, which documents the inversion presets and behaviors. Reference: ComfyUi-Untwisting-RoPE.

UntwistingRoPE (#623)

Role: injects the learned style into Krea 2 across selected U-Net blocks while compensating for rotary positional effects so structure stays novel. Tune the block range to widen or narrow where the style applies, adjust the scale parameters to balance adherence versus freedom, and use the adaptive instance normalization control to lift or tame palette and material transfer. Small key subspace alignment values can steady fine detail if edges feel jittery. Reference: ComfyUi-Untwisting-RoPE.

ImageScaleToTotalPixelsX (#265)

Role: scales the style reference to a target pixel budget and aspect ratio before encoding. Choose a resampling method like Lanczos for crisp edges, and supply dimensions that match your intended generation size. This alignment reduces aliasing in the VAE and improves the stability of the style code. Reference: ComfyUi-Scale-Image-to-Total-Pixels-Advanced.

KSampler (#635)

Role: performs the diffusion trajectory using the style-attached model and your prompt conditioning. Samplers that add noise along the path tend to wash out residual layout hints and deliver cleaner style transfer; try er_sde or euler_ancestral. Moderate steps and a balanced guidance scale usually keep both style and content on track while preserving variety through the seed.

Notes

Krea 2 style transfer ComfyUI workflow | Reference to Scene Stylization - see RunComfy page for the latest node requirements.