Portrait Relight
Upload your target image and a reference lighting image. Relights your target image.
About
The 'Portrait Relight' workflow in ComfyUI is designed to enhance your images by altering their lighting based on a reference image. This workflow utilizes the Qwen-Image-Edit model to intelligently adjust the lighting of a target image, making it appear as if it was lit under the conditions of the reference image.
The process involves several key nodes, including the LoadImage node for importing images, and the SaveImage node for exporting the final result. The workflow is particularly useful for photographers and digital artists looking to achieve consistent lighting across different images or to experiment with creative lighting effects without the need for physical lighting equipment.
What you provide
- Reading the workflow
Runs in ComfyUI. This workflow can't run on this page. Open it in Comfy Cloud with everything loaded, or download it for your own ComfyUI.
Try in CloudInside the workflow
A read-only preview of the node graph. Open it in Comfy Cloud or download it to edit the nodes.
Loading the graph
Open full-size workflow preview- Author
- Rob
Requires custom nodes
- Reading the workflow
Use this workflow from code
Generated from the workflow graph and inputs when possible; stored examples are shown otherwise. Review credentials and endpoints before running code, which may incur charges.
What you need
- Runs on
- Comfy Cloud
- Key
- COMFY_API_KEY
Needs a paid Cloud plan and credits.
# Python 3.10+: pip install comfy-sdk==0.4.0
import os
from comfy_sdk import Comfy
api_key = os.environ.get("COMFY_API_KEY", "").strip()
if not api_key: raise RuntimeError("Set COMFY_API_KEY")
os.environ.setdefault("COMFY_BASE_URL", "https://cloud.comfy.org")
client = Comfy(api_key=api_key)
workflow = client.workflows.from_json({
"5": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "ComfyUI-light_migration",
"images": [
"1:76",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"6": {
"class_type": "LoadImage",
"inputs": {
"image": "input_2-relight.png"
},
"_meta": {
"title": "Load Lighting Image"
}
},
"7": {
"class_type": "LoadImage",
"inputs": {
"image": "input_1-relight.png"
},
"_meta": {
"title": "Load Subject Image"
}
},
"1:101": {
"class_type": "LoraLoaderModelOnly",
"inputs": {
"lora_name": "Qwen-Image-Edit-2509-Light-Migration.safetensors",
"model": [
"1:89",
0
],
"strength_model": 1
},
"_meta": {
"title": "Load LoRA"
}
},
"1:102": {
"class_type": "TextEncodeQwenImageEditPlus",
"inputs": {
"clip": [
"1:38",
0
],
"image1": [
"1:104",
0
],
"image2": [
"1:105",
0
],
"prompt": "",
"vae": [
"1:39",
0
]
},
"_meta": {
"title": "Text Encode Qwen Image Edit Plus"
}
},
"1:103": {
"class_type": "TextEncodeQwenImageEditPlus",
"inputs": {
"clip": [
"1:38",
0
],
"image1": [
"1:104",
0
],
"image2": [
"1:105",
0
],
"prompt": "参考色调，移除图1原有的光照并参考图2的光照和色调对图1重新照明",
"vae": [
"1:39",
0
]
},
"_meta": {
"title": "Text Encode Qwen Image Edit Plus"
}
},
"1:104": {
"class_type": "ImageScaleToTotalPixels",
"inputs": {
"image": [
"7",
0
],
"megapixels": 1,
"resolution_steps": 1,
"upscale_method": "lanczos"
},
"_meta": {
"title": "Scale Image to Total Pixels"
}
},
"1:105": {
"class_type": "ImageScaleToTotalPixels",
"inputs": {
"image": [
"6",
0
],
"megapixels": 1,
"resolution_steps": 1,
"upscale_method": "nearest-exact"
},
"_meta": {
"title": "Scale Image to Total Pixels"
}
},
"1:108": {
"class_type": "RepeatLatentBatch",
"inputs": {
"amount": 1,
"samples": [
"1:88",
0
]
},
"_meta": {
"title": "Repeat Latent Batch"
}
},
"1:3": {
"class_type": "KSampler",
"inputs": {
"cfg": 1,
"denoise": 1,
"latent_image": [
"1:108",
0
],
"model": [
"1:75",
0
],
"negative": [
"1:103",
0
],
"positive": [
"1:102",
0
],
"sampler_name": "euler",
"scheduler": "simple",
"seed": 260787325020957,
"steps": 8
},
"_meta": {
"title": "KSampler"
}
},
"1:37": {
"class_type": "UNETLoader",
"inputs": {
"unet_name": "qwen_image_edit_2509_fp8_e4m3fn.safetensors",
"weight_dtype": "default"
},
"_meta": {
"title": "Load Diffusion Model"
}
},
"1:38": {
"class_type": "CLIPLoader",
"inputs": {
"clip_name": "qwen_2.5_vl_7b_fp8_scaled.safetensors",
"device": "default",
"type": "qwen_image"
},
"_meta": {
"title": "Load CLIP"
}
},
"1:39": {
"class_type": "VAELoader",
"inputs": {
"vae_name": "qwen_image_vae.safetensors"
},
"_meta": {
"title": "Load VAE"
}
},
"1:66": {
"class_type": "ModelSamplingAuraFlow",
"inputs": {
"model": [
"1:101",
0
],
"sampling": "flow",
"shift": 3
},
"_meta": {
"title": "ModelSamplingAuraFlow"
}
},
"1:75": {
"class_type": "CFGNorm",
"inputs": {
"model": [
"1:66",
0
],
"pre_cfg": False,
"strength": 1
},
"_meta": {
"title": "CFGNorm"
}
},
"1:76": {
"class_type": "VAEDecode",
"inputs": {
"samples": [
"1:3",
0
],
"vae": [
"1:39",
0
]
},
"_meta": {
"title": "VAE Decode"
}
},
"1:88": {
"class_type": "VAEEncode",
"inputs": {
"pixels": [
"1:104",
0
],
"vae": [
"1:39",
0
]
},
"_meta": {
"title": "VAE Encode"
}
},
"1:89": {
"class_type": "LoraLoaderModelOnly",
"inputs": {
"lora_name": "Qwen-Image-Edit-2509-Lightning-8steps-V1.0-bf16.safetensors",
"model": [
"1:37",
0
],
"strength_model": 1
},
"_meta": {
"title": "Load LoRA"
}
}
})
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)Questions & answers
Portrait images with clear lighting and shadow details tend to work best, as the model can effectively adjust these elements to match the reference lighting.
Yes, while optimized for portraits, the workflow can be applied to any image where lighting adjustments are desired.
No special hardware is required, but a GPU will significantly speed up the processing time, especially for high-resolution images.
You can append additional image edit prompt instructions in the workflow to maintain specific details, such as edges, during the relighting process.