AIGODLIKE-COMFYUI-TRANSLATION/zh-CN/Nodes/ComfyUI-Marigold.json

{
    "MarigoldModelLoader": {
        "title": "Marigold模型加载器",
        "widgets": {
            "model": "模型"
        },
        "outputs": {
            "marigold_model": "Marigold模型"
        },
        "description": "\n基于扩散的单目深度推算: \nhttps://github.com/prs-eth/Marigold  \n  \n使用 Diffusers 0.28.0 Marigold 管线。模型会自动下载到 ComfyUI/Models/diffusers 文件夹内"
    },
    "MarigoldDepthEstimation_v2": {
        "title": "Marigold_v2深度推算",
        "inputs": {
            "marigold_model": "模型",
            "image": "图像"
        },
        "widgets": {
            "seed": "随机种",
            "control_before_generate": "运行前操作",
            "denoise_steps": "步数",
            "n_repeat": "重复次数",
            "ensemble_size": "整体尺寸",
            "processing_resolution": "处理分辨率",
            "scheduler": "调度器",
            "use_taesd_vae": "使用TAESD_VAE"
        },
        "outputs": {
            "image": "图像"
        },
        "description": "\n基于扩散的单目深度推算: \nhttps://github.com/prs-eth/Marigold  \n  \n使用 Diffusers 0.28.0 Marigold 管线。"
    },
    "MarigoldDepthEstimation_v2_Video": {
        "title": "Marigold_v2深度推算视频",
        "inputs": {
            "marigold_model": "Marigold模型",
            "image": "图像"
        },
        "widgets": {
            "seed": "随机种",
            "control_before_generate": "运行前操作",
            "denoise_steps": "步数",
            "processing_resolution": "处理分辨率",
            "scheduler": "调度器",
            "blend_factor": "混合系数",
            "use_taesd_vae": "使用TAESD_VAE"
        },
        "outputs": {
            "image": "图像"
        },
        "description": "\n基于扩散的单目深度推算: \nhttps://github.com/prs-eth/Marigold  \n  \n使用 Diffusers 0.28.0 Marigold 管线。\n这个节点使用上一帧作为初始Latent，用以平滑视频。"
    },
    "MarigoldDepthEstimation": {
        "title": "Marigold深度推算",
        "inputs": {
            "image": "图像"
        },
        "widgets": {
            "seed": "随机种",
            "control_before_generate": "运行前操作",
            "denoise_steps": "步数",
            "n_repeat": "重复次数",
            "regularizer_strength": "规格化强度",
            "reduction_method": "限制方法",
            "max_iter": "最大迭代数",
            "tol": "容错值",
            "invert": "反转",
            "keep_model_loaded": "保持模型加载",
            "n_repeat_batch_size": "重复次数批次",
            "use_fp16": "使用fp16",
            "scheduler": "调度器",
            "normalize": "规格化",
            "model": "模型"
        },
        "outputs": {
            "ensembled_image": "图像"
        },
        "description": "\n基于扩散的单目深度推算: https://github.com/prs-eth/Marigold  \n  \n- 步数：每个深度图的步数，提升该值会用时间换取准确度\n- 重复次数：每个深度图的迭代次数\n- 重复次数批次：每个批次中的重复次数值, 如果你有足够的VRAM可以将该值设为与重复次数相同以提升速度  \n- 模型：Marigold 或它的 LCM 版本 marigold-lcm-v1-0，LCM模型需要约 4 步数并使用LCM调度器  \n- 调度器：不同的调度算法会得出不同结果  \n- 反转：marigold默认以黑色为近景，使用ControlNet或其他工具时需要反转  \n- 规格化强度，限制方法，最大迭代数，容错值 是用于整体流程的设置，通常情况下不要更改 \n- 使用fp16：设置为真时使用 fp16，设置为否时使用 fp32，fp16 需求VRAM更少但可能会降低图像质量 \n"
    },
    "MarigoldDepthEstimationVideo": {
        "title": "Marigold深度推算视频",
        "inputs": {
            "image": "图像"
        },
        "widgets": {
            "seed": "随机种",
            "control_before_generate": "运行前操作",
            "denoise_steps": "步数",
            "first_frame_denoise_steps": "第一帧降噪步数",
            "first_frame_n_repeat": "第一帧重复次数",
            "n_repeat": "重复次数",
            "regularizer_strength": "规格化强度",
            "reduction_method": "限制方法",
            "max_iter": "最大迭代数",
            "tol": "容错值",
            "invert": "反转",
            "keep_model_loaded": "保持模型加载",
            "n_repeat_batch_size": "重复次数批次",
            "use_fp16": "使用fp16",
            "scheduler": "调度器",
            "normalize": "规格化",
            "model": "模型"
        },
        "outputs": {
            "ensembled_image": "图像"
        },
        "description": "\n基于扩散的单目深度推算: https://github.com/prs-eth/Marigold  \n\n这个节点是用于维持序列帧一致性包含光流的实验性节点。  \n\n- 步数：每个深度图的步数，提升该值会用时间换取准确度\n- 重复次数：每个深度图的迭代次数\n- 重复次数批次：每个批次中的重复次数值, 如果你有足够的VRAM可以将该值设为与重复次数相同以提升速度  \n-  模型：Marigold 或它的 LCM 版本 marigold-lcm-v1-0，LCM模型需要约 4 步数并使用LCM调度器  \n- 调度器：不同的调度算法会得出不同结果  \n- 反转：marigold默认以黑色为近景，使用ControlNet或其他工具时需要反转  \n- 规格化强度，限制方法，最大迭代数，容错值 是用于整体流程的设置，通常情况下不要更改  \n"
    },
    "ColorizeDepthmap": {
        "title": "深度图上色",
        "inputs": {
            "image": "图像"
        },
        "widgets": {
            "colorize_method": "上色方法"
        },
        "outputs": {
            "image": "图像"
        }
    },
    "SaveImageOpenEXR": {
        "title": "保存图像为EXR",
        "inputs": {
            "images": "图像"
        },
        "widgets": {
            "filename_prefix": "文件名前缀"
        },
        "outputs": {
            "file_url": "文件URL"
        }
    },
    "RemapDepth": {
        "title": "重映射深度",
        "inputs": {
            "image": "图像"
        },
        "widgets": {
            "min": "最小值",
            "max": "最大值",
            "clamp": "钳制"
        },
        "outputs": {
            "IMAGE": "图像"
        }
    }
}