---
title: "Qwen Image Image-to-Image Refinement in ComfyUI — Node Wiring & Settings"
description: "Complete ComfyUI workflow route for Qwen Image paired with Image-to-Image Refinement (restyling an input image at 0.45–0.65 denoise). Loader: Load Diffusion Model + CLIPLoader, VRAM target: 8 GB–16 GB."
canonical: https://comfy.ink/routes/qwen-image-img2img-refiner
updated: 2026-10-17
---

# Qwen Image: Image-to-Image Refinement in ComfyUI

Blueprint for restyling an input image at 0.45–0.65 denoise with Qwen Image (8 GB–16 GB VRAM) using Load Image → VAE Encode → KSampler (denoise 0.55) → VAE Decode.

## Facts (2026-10-17)

- **Model Architecture:** Qwen Image
- **Workflow Task:** Image-to-Image Refinement
- **VRAM Requirement:** 8 GB–16 GB
- **Primary Loader:** Load Diffusion Model + CLIPLoader
- **Model Directory:** models/diffusion_models/

## Node graph blueprint: Qwen Image × Image-to-Image Refinement

This route pairs **Qwen Image** with a **Image-to-Image Refinement** setup designed for restyling an input image at 0.45–0.65 denoise. Place your model weights in `models/diffusion_models/` and start the graph with **Load Diffusion Model + CLIPLoader**.

Core wiring chain: `Load Image → VAE Encode → KSampler (denoise 0.55) → VAE Decode`. Connect your positive and negative conditioning from [CLIP Text Encode](https://comfy.ink/nodes/clip-text-encode) into [KSampler](https://comfy.ink/nodes/ksampler) and decode to RGB pixels.

## VRAM tuning and execution tips on Windows

- **Memory target (8 GB–16 GB):** If your GPU sits at 8 GB or 12 GB, use FP8 or GGUF Q4/Q5 weights and swap in [VAE Decode (Tiled)](https://comfy.ink/nodes/vae-decode-encode) to avoid end-of-pass memory spikes.
- **Workflow persistence:** Every image saved by this route embeds the full JSON graph inside the PNG metadata — drag the output back onto [ComfyUI Studio](https://comfy.ink/studio) anytime to resume.
- **Get the app:** Download the Windows desktop or portable build from [/download](https://comfy.ink/download).

## FAQ

### Which loader node do I use for Qwen Image in a Image-to-Image Refinement?

Use `Load Diffusion Model + CLIPLoader` with weights stored in `models/diffusion_models/`, then wire the output into `Load Image → VAE Encode → KSampler (denoise 0.55) → VAE Decode`.

### How much VRAM does Qwen Image need for Image-to-Image Refinement?

Target VRAM is 8 GB–16 GB. On 8 GB cards, load quantized FP8 or GGUF weights and enable tiled VAE decoding.

## Related

- [ComfyUI Model & Task Routes](https://comfy.ink/routes)
- [ComfyUI Models & Checkpoints](https://comfy.ink/models)
- [KSampler & KSampler (Advanced)](https://comfy.ink/nodes/ksampler)
- [ComfyUI VRAM Requirements for SDXL, Flux & Video (6GB–24GB)](https://comfy.ink/guides/comfyui-vram-requirements-guide)
- [Flux.1: Text-to-Image Graph in ComfyUI](https://comfy.ink/routes/flux-txt2img-workflow)

## Download

Free desktop app: [Download Comfy](https://comfy.ink/download)

Markdown mirror: https://comfy.ink/routes/qwen-image-img2img-refiner.md