stable diffusion 局部重绘 reference-only api 接口调试

webUI api payload 插件生成的接口参数不准确,reference-only 的image不是对象,就是不同字符串字段,直接传,不是套image。

综上,那个插件参数不确定,应直接看插件的源码,看它接受什么参数

复制代码
错误:"image":
    {
        "image": None,
        "mask": None
    },

正确 "image":xxxxxx

看 stable diffusion 控制台输出,说用上了reference-only

如果 image 传岔劈了,控制台不会提示用了这个插件

请求参数-局部重绘

{
    "prompt": "",
    "negative_prompt": "",
    "batch_size": 1,
    "n_iter": 2,  # "n_iter": 1,
    "seed": "-1",
    "steps": 30,
    "cfg_scale": 7,
    "sampler_name": "DPM++ 2M Karras",
    "width": 768,
    "height": 1152,  # fact_num
    "init_images": [],
    # 控制图像变化的程度,如果=0,没有任何变化,如果=1,则新图像不会跟随输入图像,0.75是个很好的起点
    "denoising_strength": 0.9,
    "resize_mode": 0,  # "resize_mode": 2,
    "mask": None,
    "inpainting_fill": 1,  # mask content
    "inpaint_full_res": 0,  # inpaint area
    "inpaint_full_res_padding": 32,
    "alwayson_scripts": {
        "ControlNet": {
            "args": [
                {
                    "advanced_weighting": None,
                    "batch_images": "",
                    "control_mode": "ControlNet is more important",
                    "enabled": True,
                    "guidance_end": 1,
                    "guidance_start": 0,
                    "hr_option": "Both",
                    "image": None,
                    #"image": 
                        #{
                            #"image": None,
                            #"mask": None
                        #},
                    "inpaint_crop_input_image": True,
                    "input_mode": "simple",
                    "is_ui": True,
                    "loopback": False,
                    "low_vram": False,
                    "model": "None",
                    "module": "reference_only",
                    "output_dir": "",
                    "pixel_perfect": True,
                    "processor_res": -1,
                    "resize_mode": "Crop and Resize",
                    "save_detected_map": True,
                    "threshold_a": 1,
                    "threshold_b": -1,
                    "weight": 2
                },
                {
                    "advanced_weighting": None,
                    "batch_images": "",
                    "control_mode": "Balanced",
                    "enabled": True,
                    "guidance_end": 1,
                    "guidance_start": 0,
                    "hr_option": "Both",
                    "image":
                        {
                            "image": None,
                            "mask": None
                        },
                    "inpaint_crop_input_image": True,
                    "input_mode": "simple",
                    "is_ui": True,
                    "loopback": False,
                    "low_vram": False,
                    "model": "control_v11p_sd15_openpose [cab727d4]",
                    "module": "openpose_full",
                    "output_dir": "",
                    "pixel_perfect": True,
                    "processor_res": 512,
                    "resize_mode": "Crop and Resize",
                    "save_detected_map": True,
                    "threshold_a": -1,
                    "threshold_b": -1,
                    "weight": 1
                }
            ]
        }
    }
}
相关推荐
yuzhangfeng2 天前
【 模型】 开源图像模型Stable Diffusion入门手册
stable diffusion
浪淘沙jkp3 天前
(一)Ubuntu22.04服务器端部署Stable-Diffusion-webui AI绘画环境
ai作画·stable diffusion
kebijuelun4 天前
阿里数字人工作 Emote Portrait Alive (EMO):基于 Diffusion 直接生成视频的数字人方案
人工智能·语言模型·stable diffusion·aigc·音视频
李歘歘6 天前
Stable Diffusion概要讲解
stable diffusion
davenian7 天前
<AI 学习> 下载 Stable Diffusions via Windows OS
学习·stable diffusion
AI极客菌10 天前
[ComfyUI]Flux:繁荣生态魔盒已开启,6款LORA已来,更有MJ6&写实&动漫&风景&艺术&迪士尼全套
ai作画·stable diffusion·aigc·midjourney·人工智能作画·comfyui·风景
许野平11 天前
Stable Diffusion:照片转视频插件
stable diffusion·animatediff·temporal·照片转视频
python零基础入门小白12 天前
如何通过AI绘画技术轻松变现?教你三种方法赚取高额利润!
ai作画·stable diffusion·aigc
AI设计酷卡14 天前
什么是 Stable Diffusion? | StableDiffusion 基础知识
人工智能·stable diffusion·aigc
AI极客菌16 天前
Controlnet作者新作IC-light V2:基于FLUX训练,支持处理风格化图像,细节远高于SD1.5。
人工智能·计算机视觉·ai作画·stable diffusion·aigc·flux·人工智能作画