Qwen-Image-2.1 Image Edit (No PE)
Reference-based image editing with Qwen-Image-2.1. Same as the official ComfyUI template, with the prompt-enhancer (PE) branch removed so it runs on hosts that lack the qwen3.5_9b PE model. Your prompt goes straight CFG 1.0.
Qwen-Image-2.1 Image Edit i2i Reference
40
Nodes & Models
MarkdownNote
LoadImage
UNETLoader
qwen_image_2.1_int8_convrot.safetensors
CLIPLoader
qwen3vl_8b_int8_convrot.safetensors
VAELoader
qwen_image_2.1_vae_bf16.safetensors
SaveImageAdvanced
EmptyLatentImage
ImageCompare
VAEDecode
KSampler
ComfySwitchNode
QwenImage21Cache
TextEncodeQwenImage21
# Qwen-Image-2.1 Image Edit (No PE)
Reference-based image editing with Qwen-Image-2.1. Based on the official ComfyUI template (image_qwen_image_2_1_image_edit) with the prompt-enhancer branch removed, so it does not require the qwen3.5_9b PE model.
## How to use
1. Load one or more reference images (up to 10). image_1 is the edit target.
2. Write the edit instruction in positive_prompt. Refer to inputs as <image1>, <image2>, etc. Japanese and other languages work.
3. Run. Output size follows image_1 unless custom_size is on.
## Settings
- Steps: 25
- CFG: 1.0
- Sampler: euler / simple
## Models required
- qwen_image_2.1_int8_convrot.safetensors (UNET)
- qwen3vl_8b_int8_convrot.safetensors (CLIP)
- qwen_image_2.1_vae_bf16.safetensors (VAE)
## License note
Qwen-Image-2.1 is released under the Qwen Research License (non-commercial). Check the license before commercial use.
Read more
