Krea 2 Turbo delivers fast local edits and preserves everything else.

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

Use Krea 2 Turbo with LanPaint to edit only what you mask. Replace objects while keeping nearby pixels intact. Let SAM3 build masks automatically. Crop-and-stitch improves local detail. Differential Diffusion adds control. You get fast, targeted image repairs.

Important nodes:

Key nodes in Krea 2 Turbo inpainting ComfyUI workflow

• LanPaint_KSampler (#440). The core training‑free inpainting step that repurposes a text‑to‑image model for localized edits without fine‑tuning. It samples only where the mask allows, guided by your positive conditioning. For stronger adherence to your instruction, increase the sampling effort; for tighter realism with the photo, bias toward conservative denoising. Implementation: scraed/LanPaint, paper: LanPaint (arXiv).

• SetLatentNoiseMask (#430). Ensures noise is applied only inside the mask in latent space so the background latents remain locked to the original image. If you see unintended changes outside the target, revisit this step and confirm your mask covers only what should change.

• DifferentialDiffusionAdvanced (#447). Constrains the update step so the unmasked region resists drift. When small halos or tonal shifts appear near edges, dial this toward stronger preservation so the context stays faithful while the masked area evolves.

• GrowMaskWithBlur (#452). Expands and feathers the mask to capture contact shadows and interreflections around the object. If seams are visible, grow slightly more and add a touch of blur; if edits spill too far, reduce the growth.

• CropByMask (#449). Crops the region of interest and records the crop origin. This keeps generation fast and focused, and it drives exact placement later. If the edit lacks enough context, add a bit more padding so the model can see adjacent surfaces and lighting cues.

• ImageCompositeMasked (#487). Rebuilds the full canvas by placing the edited patch back at its original position using the saved coordinates. If you notice edge softness or misalignment, confirm the crop origin and that the stitching mask matches the grown mask from the generation phase.

• SAM3Segment (#454). Produces an automatic segmentation mask from the input image. It is handy for fast starts; refine with a brush if needed to tighten selection around tricky boundaries like hair, spokes, or filigree.

Notes

Krea 2 Turbo Inpainting in ComfyUI | LanPaint + SAM3 - see RunComfy page for the latest node requirements.