Flux.1 Dev OneReward
Supports various tasks such as image inpainting, outpainting, and object removal by bytedance-research team
概要
The 'Flux.1 Dev OneReward' workflow is a versatile tool designed for image manipulation tasks such as inpainting, outpainting, and object removal. Developed by the bytedance-research team, this workflow leverages the Flux.1 model to seamlessly fill in or extend parts of an image. The workflow employs a series of nodes, including LoadImage, SaveImage, and PreviewImage, to manage the input and output processes. The core functionality is driven by nodes like cb0eaf1c-704f-477d-8893-79665db14ed1 and b8560576-5524-4495-baa5-2cb40da12e9e, which handle the complex computations required for high-quality image editing.
This makes it particularly useful for creative professionals and enthusiasts who need precise control over image content.
入力するもの
- ワークフローを読み込み中
ComfyUI で実行できます。 このワークフローはこのページでは実行できません。すべてが読み込まれた状態でComfy Cloudで開くか、ご自身のComfyUI用にダウンロードしてください。
Cloudで試すワークフローの内部
ノードグラフの読み取り専用プレビューです。ノードを編集するには、Comfy Cloudで開くかダウンロードしてください。
Loading the graph
Open full-size workflow preview- 作成者
- ComfyUI
カスタムノードが必要です
- ワークフローを読み込み中
コードからこのワークフローを使用する
可能な場合はワークフローグラフと入力から生成されます。それ以外の場合は保存された例が表示されます。料金が発生する場合があるため、コードを実行する前に認証情報とエンドポイントを確認してください。
What you need
- Runs on
- Comfy Cloud
- キー
- COMFY_API_KEY
有料の Cloud プランとクレジットが必要です。
# 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({
"17": {
"class_type": "LoadImage",
"inputs": {
"image": "image_flux.1_fill_dev_OneReward_input_image.png"
},
"_meta": {
"title": "Load Image"
}
},
"62": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "ComfyUI",
"images": [
"110:65",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"105": {
"class_type": "PreviewImage",
"inputs": {
"images": [
"108:8",
0
]
},
"_meta": {
"title": "Preview Image"
}
},
"121": {
"class_type": "PreviewImage",
"inputs": {
"images": [
"17",
0
]
},
"_meta": {
"title": "Preview Image"
}
},
"108:100": {
"class_type": "LoraLoaderModelOnly",
"inputs": {
"lora_name": "removal_timestep_alpha-2-1740.safetensors",
"model": [
"108:31",
0
],
"strength_model": 1
},
"_meta": {
"title": "Load LoRA"
}
},
"108:23": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"108:34",
0
],
"text": "Remove the girl's hat\n"
},
"_meta": {
"title": "CLIP Text Encode (Positive Prompt)"
}
},
"108:26": {
"class_type": "FluxGuidance",
"inputs": {
"conditioning": [
"108:23",
0
],
"guidance": 30
},
"_meta": {
"title": "FluxGuidance"
}
},
"108:3": {
"class_type": "KSampler",
"inputs": {
"cfg": 1,
"denoise": 1,
"latent_image": [
"108:38",
2
],
"model": [
"108:39",
0
],
"negative": [
"108:38",
1
],
"positive": [
"108:38",
0
],
"sampler_name": "euler",
"scheduler": "normal",
"seed": 1100974460616450,
"steps": 20
},
"_meta": {
"title": "KSampler"
}
},
"108:31": {
"class_type": "UNETLoader",
"inputs": {
"unet_name": "flux.1-fill-dev-OneReward-transformer_fp8.safetensors",
"weight_dtype": "default"
},
"_meta": {
"title": "Load Diffusion Model"
}
},
"108:32": {
"class_type": "VAELoader",
"inputs": {
"vae_name": "ae.safetensors"
},
"_meta": {
"title": "Load VAE"
}
},
"108:34": {
"class_type": "DualCLIPLoader",
"inputs": {
"clip_name1": "clip_l.safetensors",
"clip_name2": "t5xxl_fp16.safetensors",
"device": "default",
"type": "flux"
},
"_meta": {
"title": "Load CLIP (Dual)"
}
},
"108:38": {
"class_type": "InpaintModelConditioning",
"inputs": {
"mask": [
"17",
1
],
"negative": [
"108:46",
0
],
"noise_mask": True,
"pixels": [
"17",
0
],
"positive": [
"108:26",
0
],
"vae": [
"108:32",
0
]
},
"_meta": {
"title": "InpaintModelConditioning"
}
},
"108:39": {
"class_type": "DifferentialDiffusion",
"inputs": {
"model": [
"108:100",
0
],
"strength": 1
},
"_meta": {
"title": "Differential Diffusion"
}
},
"108:46": {
"class_type": "ConditioningZeroOut",
"inputs": {
"conditioning": [
"108:23",
0
]
},
"_meta": {
"title": "Conditioning Zero Out"
}
},
"108:8": {
"class_type": "VAEDecode",
"inputs": {
"samples": [
"108:3",
0
],
"vae": [
"108:32",
0
]
},
"_meta": {
"title": "VAE Decode"
}
},
"110:57": {
"class_type": "DualCLIPLoader",
"inputs": {
"clip_name1": "clip_l.safetensors",
"clip_name2": "t5xxl_fp16.safetensors",
"device": "default",
"type": "flux"
},
"_meta": {
"title": "Load CLIP (Dual)"
}
},
"110:58": {
"class_type": "KSampler",
"inputs": {
"cfg": 1,
"denoise": 1,
"latent_image": [
"110:73",
0
],
"model": [
"110:63",
0
],
"negative": [
"110:71",
1
],
"positive": [
"110:71",
0
],
"sampler_name": "euler",
"scheduler": "normal",
"seed": 325184035774450,
"steps": 20
},
"_meta": {
"title": "KSampler"
}
},
"110:59": {
"class_type": "VAELoader",
"inputs": {
"vae_name": "ae.safetensors"
},
"_meta": {
"title": "Load VAE"
}
},
"110:60": {
"class_type": "FluxGuidance",
"inputs": {
"conditioning": [
"110:74",
0
],
"guidance": 30
},
"_meta": {
"title": "FluxGuidance"
}
},
"110:61": {
"class_type": "ConditioningZeroOut",
"inputs": {
"conditioning": [
"110:74",
0
]
},
"_meta": {
"title": "Conditioning Zero Out"
}
},
"110:63": {
"class_type": "DifferentialDiffusion",
"inputs": {
"model": [
"110:64",
0
],
"strength": 1
},
"_meta": {
"title": "Differential Diffusion"
}
},
"110:64": {
"class_type": "UNETLoader",
"inputs": {
"unet_name": "flux.1-fill-dev-OneReward-transformer_fp8.safetensors",
"weight_dtype": "default"
},
"_meta": {
"title": "Load Diffusion Model"
}
},
"110:65": {
"class_type": "VAEDecode",
"inputs": {
"samples": [
"110:58",
0
],
"vae": [
"110:59",
0
]
},
"_meta": {
"title": "VAE Decode"
}
},
"110:71": {
"class_type": "InpaintModelConditioning",
"inputs": {
"mask": [
"110:75",
1
],
"negative": [
"110:61",
0
],
"noise_mask": False,
"pixels": [
"110:75",
0
],
"positive": [
"110:60",
0
],
"vae": [
"110:59",
0
]
},
"_meta": {
"title": "InpaintModelConditioning"
}
},
"110:73": {
"class_type": "SetLatentNoiseMask",
"inputs": {
"mask": [
"110:75",
1
],
"samples": [
"110:71",
2
]
},
"_meta": {
"title": "Set Latent Noise Mask"
}
},
"110:74": {
"class_type": "CLIPTextEncode",
"inputs": {
"clip": [
"110:57",
0
],
"text": ""
},
"_meta": {
"title": "CLIP Text Encode (Positive Prompt)"
}
},
"110:75": {
"class_type": "ImagePadForOutpaint",
"inputs": {
"bottom": 104,
"feathering": 40,
"image": [
"17",
0
],
"left": 40,
"right": 104,
"top": 0
},
"_meta": {
"title": "Pad Image for Outpainting"
}
}
})
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)よくある質問
The Flux.1 model is used for inpainting and outpainting tasks, allowing for seamless filling or extension of image areas.
No, this workflow is specifically designed for still images. Video editing would require a different set of tools and models.
You can edit the mask by right-clicking on the Load Image node and selecting Open in 'Maskeditor' to define the areas you want to modify.
Yes, it is recommended to update ComfyUI to ensure compatibility and access to the latest features and models.