ComfyUI - 使用 ComfyUI 部署与测试 FLUX.1 图像生成模型 教程

已于 2024-08-15 10:23:34 修改
阅读量433 收藏 4
点赞数 13
分类专栏： Stable Diffusion 文章标签： Flux ComfyUI CLIP HuggingFace AIGC
于 2024-08-15 01:37:54 首次发布
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/caroline_wendy/article/details/141201307
版权

欢迎关注我的CSDN：https://spike.blog.csdn.net/
本文地址：https://spike.blog.csdn.net/article/details/141201307

免责声明：本文来源于个人知识与公开资料，仅用于学术交流，欢迎讨论，不支持转载。

FLUX.1 是由 Black Forest Labs 推出的文本到图像生成模型，已成为 AI 绘画领域的高品质模型。该模型由 Stability AI 的前核心成员开发，具备强大的生成能力和高质量的图像输出。目前，Flux 的相关模型：

  • Flux & AE 模型：https://huggingface.co/black-forest-labs/FLUX.1-dev
  • CLIP 模型：https://huggingface.co/stabilityai/stable-diffusion-3-medium

FP8 模型地址：https://huggingface.co/Kijai/flux-fp8/tree/main

安装 HuggingFace 下载工具，使用镜像下载速度明显加快：

export HF_ENDPOINT="https://hf-mirror.com"
pip install -U huggingface_hub hf-transfer

下载 HuggingFace 脚本，如下：

huggingface-cli download --token [your toekn] black-forest-labs/FLUX.1-dev --local-dir FLUX.1-dev --include "flux1-dev.safetensors" 
huggingface-cli download --token [your toekn] black-forest-labs/FLUX.1-dev --local-dir FLUX.1-dev --include "ae.safetensors"
huggingface-cli download --token [your toekn] stabilityai/stable-diffusion-3-medium --local-dir stable-diffusion-3-medium

下载之前需要申请权限，Token 地址来源于：https://huggingface.co/settings/tokens，全部勾选即可生成。

其中，完全版是 FP16 的版本 flux1-dev.safetensors，大约 23G，即：

FLUX.1-dev/flux1-dev.safetensors 
bypy upload flux1-dev.safetensors /stable_diffusion/flux_data/flux1-dev.safetensors

AE 模型：

FLUX.1-dev/ae.safetensors

CLIP 是两个模型：

stable-diffusion-3-medium/text_encoders/clip_l.safetensors
stable-diffusion-3-medium/text_encoders/t5xxl_fp16.safetensors

将 Flux 的相关模型，放入相应的位置：

ComfyUI/models/unet/flux1-dev.safetensors
ComfyUI/models/vae/ae.safetensors
ComfyUI/models/clip/clip_l.safetensors
ComfyUI/models/clip/t5xxl_fp16.safetensors

搭建 Flux 流程：

Pipeline

相关提示词：

A queen sits on a throne, looking at the camera with arrogance, legs crossed, wearing high heels and white stockings.
The palace is bright, highlighting the queen’s majesty.
Low POV (Point of View), tilted shot from below.

Image

A Chinese queen sits on a very high throne in bright palace, looking at the camera, legs crossed, wearing golden high heels and black stockings.
very low POV, tilted shot from below.

Image

A woman empress of China sits on a very high throne in the bright palace. She is looking at the camera, legs crossed, wearing golden high heels and black stockings.
very low POV, tilted shot from very below.
very high quality.

Image
原图：
Image
重绘图像：
Image
ComfyUI Pipeline Json

{
  "last_node_id": 66,
  "last_link_id": 110,
  "nodes": [
    {
      "id": 8,
      "type": "UNETLoader",
      "pos": [
        -372,
        -138
      ],
      "size": {
        "0": 315,
        "1": 82
      },
      "flags": {},
      "order": 0,
      "mode": 0,
      "outputs": [
        {
          "name": "MODEL",
          "type": "MODEL",
          "links": [
            73
          ],
          "shape": 3,
          "label": "模型",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "UNETLoader"
      },
      "widgets_values": [
        "flux1-dev.safetensors",
        "default"
      ]
    },
    {
      "id": 9,
      "type": "DualCLIPLoader",
      "pos": [
        -372,
        164
      ],
      "size": {
        "0": 315,
        "1": 106
      },
      "flags": {},
      "order": 1,
      "mode": 0,
      "outputs": [
        {
          "name": "CLIP",
          "type": "CLIP",
          "links": [
            81
          ],
          "shape": 3,
          "label": "CLIP",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "DualCLIPLoader"
      },
      "widgets_values": [
        "t5xxl_fp16.safetensors",
        "clip_l.safetensors",
        "flux"
      ]
    },
    {
      "id": 12,
      "type": "CLIPTextEncodeFlux",
      "pos": [
        5,
        165
      ],
      "size": {
        "0": 246.04954528808594,
        "1": 111.51980590820312
      },
      "flags": {},
      "order": 11,
      "mode": 0,
      "inputs": [
        {
          "name": "clip",
          "type": "CLIP",
          "link": 81,
          "label": "CLIP"
        },
        {
          "name": "clip_l",
          "type": "STRING",
          "link": 34,
          "widget": {
            "name": "clip_l"
          },
          "label": "CLIP_L",
          "slot_index": 1
        },
        {
          "name": "t5xxl",
          "type": "STRING",
          "link": 35,
          "widget": {
            "name": "t5xxl"
          },
          "label": "T5XXL"
        }
      ],
      "outputs": [
        {
          "name": "CONDITIONING",
          "type": "CONDITIONING",
          "links": [
            14
          ],
          "shape": 3,
          "label": "条件",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "CLIPTextEncodeFlux"
      },
      "widgets_values": [
        "A queen sits on a throne, looking at the camera with arrogance, legs crossed, wearing high heels and black stockings.\nLow POV, tilted shot from below.\nThe palace is bathed in natural light.",
        "A queen sits on a throne, looking at the camera with arrogance, legs crossed, wearing high heels and black stockings.\nLow POV, tilted shot from below.\nThe palace is bathed in natural light.",
        3.5,
        true,
        true
      ]
    },
    {
      "id": 15,
      "type": "LatentFromBatch",
      "pos": [
        317,
        135
      ],
      "size": {
        "0": 315,
        "1": 82
      },
      "flags": {},
      "order": 9,
      "mode": 0,
      "inputs": [
        {
          "name": "samples",
          "type": "LATENT",
          "link": 13,
          "label": "Latent"
        }
      ],
      "outputs": [
        {
          "name": "LATENT",
          "type": "LATENT",
          "links": [
            18
          ],
          "shape": 3,
          "label": "Latent",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "LatentFromBatch"
      },
      "widgets_values": [
        0,
        1
      ]
    },
    {
      "id": 5,
      "type": "EmptyLatentImage",
      "pos": [
        -372,
        -291
      ],
      "size": {
        "0": 315,
        "1": 106
      },
      "flags": {},
      "order": 2,
      "mode": 0,
      "outputs": [
        {
          "name": "LATENT",
          "type": "LATENT",
          "links": [
            13
          ],
          "shape": 3,
          "label": "Latent",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "EmptyLatentImage"
      },
      "widgets_values": [
        1280,
        800,
        1
      ]
    },
    {
      "id": 16,
      "type": "BasicScheduler",
      "pos": [
        317,
        279
      ],
      "size": {
        "0": 315,
        "1": 106
      },
      "flags": {},
      "order": 12,
      "mode": 0,
      "inputs": [
        {
          "name": "model",
          "type": "MODEL",
          "link": 79,
          "label": "模型"
        }
      ],
      "outputs": [
        {
          "name": "SIGMAS",
          "type": "SIGMAS",
          "links": [
            20
          ],
          "shape": 3,
          "label": "Sigmas",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "BasicScheduler"
      },
      "widgets_values": [
        "normal",
        25,
        1
      ]
    },
    {
      "id": 11,
      "type": "ModelSamplingFlux",
      "pos": [
        -372,
        -10
      ],
      "size": {
        "0": 315,
        "1": 130
      },
      "flags": {},
      "order": 8,
      "mode": 0,
      "inputs": [
        {
          "name": "model",
          "type": "MODEL",
          "link": 73,
          "label": "模型"
        }
      ],
      "outputs": [
        {
          "name": "MODEL",
          "type": "MODEL",
          "links": [
            78,
            79,
            99
          ],
          "shape": 3,
          "label": "模型",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "ModelSamplingFlux"
      },
      "widgets_values": [
        1.15,
        0.5,
        1024,
        1024
      ]
    },
    {
      "id": 61,
      "type": "Reroute",
      "pos": [
        997,
        -64
      ],
      "size": [
        75,
        26
      ],
      "flags": {},
      "order": 13,
      "mode": 0,
      "inputs": [
        {
          "name": "",
          "type": "*",
          "link": 99
        }
      ],
      "outputs": [
        {
          "name": "",
          "type": "MODEL",
          "links": [
            100
          ],
          "slot_index": 0
        }
      ],
      "properties": {
        "showOutputText": false,
        "horizontal": false
      }
    },
    {
      "id": 17,
      "type": "BasicGuider",
      "pos": [
        317,
        -58
      ],
      "size": {
        "0": 241.79998779296875,
        "1": 46
      },
      "flags": {},
      "order": 14,
      "mode": 0,
      "inputs": [
        {
          "name": "model",
          "type": "MODEL",
          "link": 78,
          "label": "模型"
        },
        {
          "name": "conditioning",
          "type": "CONDITIONING",
          "link": 14,
          "label": "条件"
        }
      ],
      "outputs": [
        {
          "name": "GUIDER",
          "type": "GUIDER",
          "links": [
            17,
            101
          ],
          "shape": 3,
          "label": "引导",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "BasicGuider"
      }
    },
    {
      "id": 14,
      "type": "KSamplerSelect",
      "pos": [
        317,
        29
      ],
      "size": {
        "0": 315,
        "1": 58
      },
      "flags": {},
      "order": 3,
      "mode": 0,
      "outputs": [
        {
          "name": "SAMPLER",
          "type": "SAMPLER",
          "links": [
            19,
            103
          ],
          "shape": 3,
          "label": "采样器",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "KSamplerSelect"
      },
      "widgets_values": [
        "euler"
      ]
    },
    {
      "id": 19,
      "type": "VAEDecode",
      "pos": [
        717,
        -64
      ],
      "size": {
        "0": 210,
        "1": 46
      },
      "flags": {
        "collapsed": false
      },
      "order": 17,
      "mode": 0,
      "inputs": [
        {
          "name": "samples",
          "type": "LATENT",
          "link": 93,
          "label": "Latent"
        },
        {
          "name": "vae",
          "type": "VAE",
          "link": 22,
          "label": "VAE"
        }
      ],
      "outputs": [
        {
          "name": "IMAGE",
          "type": "IMAGE",
          "links": [
            89
          ],
          "shape": 3,
          "label": "图像",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "VAEDecode"
      }
    },
    {
      "id": 10,
      "type": "VAELoader",
      "pos": [
        -372,
        321
      ],
      "size": {
        "0": 315,
        "1": 58
      },
      "flags": {},
      "order": 4,
      "mode": 0,
      "outputs": [
        {
          "name": "VAE",
          "type": "VAE",
          "links": [
            22,
            104
          ],
          "shape": 3,
          "slot_index": 0,
          "label": "VAE"
        }
      ],
      "properties": {
        "Node name for S&R": "VAELoader"
      },
      "widgets_values": [
        "flux_ae.safetensors"
      ]
    },
    {
      "id": 63,
      "type": "Reroute",
      "pos": [
        999,
        -33
      ],
      "size": [
        75,
        26
      ],
      "flags": {},
      "order": 10,
      "mode": 0,
      "inputs": [
        {
          "name": "",
          "type": "*",
          "link": 104
        }
      ],
      "outputs": [
        {
          "name": "",
          "type": "VAE",
          "links": [
            105
          ],
          "slot_index": 0
        }
      ],
      "properties": {
        "showOutputText": false,
        "horizontal": false
      }
    },
    {
      "id": 62,
      "type": "VAEDecode",
      "pos": [
        1135,
        92
      ],
      "size": {
        "0": 210,
        "1": 46
      },
      "flags": {
        "collapsed": false
      },
      "order": 21,
      "mode": 0,
      "inputs": [
        {
          "name": "samples",
          "type": "LATENT",
          "link": 106,
          "label": "Latent"
        },
        {
          "name": "vae",
          "type": "VAE",
          "link": 105,
          "label": "VAE"
        }
      ],
      "outputs": [
        {
          "name": "IMAGE",
          "type": "IMAGE",
          "links": [
            107
          ],
          "shape": 3,
          "label": "图像",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "VAEDecode"
      }
    },
    {
      "id": 66,
      "type": "RandomNoise",
      "pos": [
        1458,
        -350
      ],
      "size": {
        "0": 315,
        "1": 82
      },
      "flags": {},
      "order": 5,
      "mode": 0,
      "outputs": [
        {
          "name": "NOISE",
          "type": "NOISE",
          "links": [
            110
          ],
          "shape": 3,
          "label": "噪波生成",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "RandomNoise"
      },
      "widgets_values": [
        858548031216041,
        "randomize"
      ]
    },
    {
      "id": 18,
      "type": "SamplerCustomAdvanced",
      "pos": [
        713,
        -219
      ],
      "size": {
        "0": 355.20001220703125,
        "1": 106
      },
      "flags": {},
      "order": 16,
      "mode": 0,
      "inputs": [
        {
          "name": "noise",
          "type": "NOISE",
          "link": 16,
          "label": "噪波生成"
        },
        {
          "name": "guider",
          "type": "GUIDER",
          "link": 17,
          "label": "引导"
        },
        {
          "name": "sampler",
          "type": "SAMPLER",
          "link": 19,
          "label": "采样器"
        },
        {
          "name": "sigmas",
          "type": "SIGMAS",
          "link": 20,
          "label": "Sigmas"
        },
        {
          "name": "latent_image",
          "type": "LATENT",
          "link": 18,
          "label": "Latent"
        }
      ],
      "outputs": [
        {
          "name": "output",
          "type": "LATENT",
          "links": [
            93,
            108
          ],
          "shape": 3,
          "label": "输出",
          "slot_index": 0
        },
        {
          "name": "denoised_output",
          "type": "LATENT",
          "links": [],
          "shape": 3,
          "label": "降噪输出",
          "slot_index": 1
        }
      ],
      "properties": {
        "Node name for S&R": "SamplerCustomAdvanced"
      },
      "color": "#322",
      "bgcolor": "#533"
    },
    {
      "id": 13,
      "type": "RandomNoise",
      "pos": [
        317,
        -186
      ],
      "size": {
        "0": 315,
        "1": 82
      },
      "flags": {},
      "order": 6,
      "mode": 0,
      "outputs": [
        {
          "name": "NOISE",
          "type": "NOISE",
          "links": [
            16
          ],
          "shape": 3,
          "label": "噪波生成",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "RandomNoise"
      },
      "widgets_values": [
        3,
        "fixed"
      ]
    },
    {
      "id": 37,
      "type": "StringFunction|pysssss",
      "pos": [
        -373,
        442
      ],
      "size": {
        "0": 403.92303466796875,
        "1": 274
      },
      "flags": {},
      "order": 7,
      "mode": 0,
      "outputs": [
        {
          "name": "STRING",
          "type": "STRING",
          "links": [
            34,
            35
          ],
          "shape": 3,
          "label": "字符串",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "StringFunction|pysssss"
      },
      "widgets_values": [
        "append",
        "yes",
        "A woman empress of China sits on a very high throne in the bright palace. She is looking at the camera, legs crossed, wearing golden high heels and black stockings.\n",
        "very low POV, tilted shot from very below.\nvery high quality.",
        "",
        "A woman empress of China sits on a very high throne in the bright palace. She is looking at the camera, legs crossed, wearing golden high heels and black stockings.\n, very low POV, tilted shot from very below.\nvery high quality."
      ]
    },
    {
      "id": 65,
      "type": "LatentUpscaleBy",
      "pos": [
        1127,
        -350
      ],
      "size": {
        "0": 315,
        "1": 82
      },
      "flags": {},
      "order": 18,
      "mode": 0,
      "inputs": [
        {
          "name": "samples",
          "type": "LATENT",
          "link": 108,
          "label": "Latent"
        }
      ],
      "outputs": [
        {
          "name": "LATENT",
          "type": "LATENT",
          "links": [
            109
          ],
          "shape": 3,
          "label": "Latent",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "LatentUpscaleBy"
      },
      "widgets_values": [
        "nearest-exact",
        1.5
      ]
    },
    {
      "id": 59,
      "type": "SamplerCustomAdvanced",
      "pos": [
        1126,
        -219
      ],
      "size": {
        "0": 355.20001220703125,
        "1": 106
      },
      "flags": {},
      "order": 20,
      "mode": 0,
      "inputs": [
        {
          "name": "noise",
          "type": "NOISE",
          "link": 110,
          "label": "噪波生成"
        },
        {
          "name": "guider",
          "type": "GUIDER",
          "link": 101,
          "label": "引导"
        },
        {
          "name": "sampler",
          "type": "SAMPLER",
          "link": 103,
          "label": "采样器"
        },
        {
          "name": "sigmas",
          "type": "SIGMAS",
          "link": 98,
          "label": "Sigmas"
        },
        {
          "name": "latent_image",
          "type": "LATENT",
          "link": 109,
          "label": "Latent"
        }
      ],
      "outputs": [
        {
          "name": "output",
          "type": "LATENT",
          "links": [
            106
          ],
          "shape": 3,
          "label": "输出",
          "slot_index": 0
        },
        {
          "name": "denoised_output",
          "type": "LATENT",
          "links": [],
          "shape": 3,
          "label": "降噪输出",
          "slot_index": 1
        }
      ],
      "properties": {
        "Node name for S&R": "SamplerCustomAdvanced"
      },
      "color": "#322",
      "bgcolor": "#533"
    },
    {
      "id": 60,
      "type": "BasicScheduler",
      "pos": [
        1130,
        -63
      ],
      "size": {
        "0": 315,
        "1": 106
      },
      "flags": {},
      "order": 15,
      "mode": 0,
      "inputs": [
        {
          "name": "model",
          "type": "MODEL",
          "link": 100,
          "label": "模型",
          "slot_index": 0
        }
      ],
      "outputs": [
        {
          "name": "SIGMAS",
          "type": "SIGMAS",
          "links": [
            98
          ],
          "shape": 3,
          "label": "Sigmas",
          "slot_index": 0
        }
      ],
      "properties": {
        "Node name for S&R": "BasicScheduler"
      },
      "widgets_values": [
        "normal",
        20,
        0.5
      ]
    },
    {
      "id": 7,
      "type": "PreviewImage",
      "pos": [
        716,
        30
      ],
      "size": {
        "0": 359.57421875,
        "1": 249.54296875
      },
      "flags": {},
      "order": 19,
      "mode": 0,
      "inputs": [
        {
          "name": "images",
          "type": "IMAGE",
          "link": 89,
          "label": "图像"
        }
      ],
      "properties": {
        "Node name for S&R": "PreviewImage"
      }
    },
    {
      "id": 64,
      "type": "PreviewImage",
      "pos": [
        1137,
        190
      ],
      "size": {
        "0": 359.57421875,
        "1": 249.54296875
      },
      "flags": {},
      "order": 22,
      "mode": 0,
      "inputs": [
        {
          "name": "images",
          "type": "IMAGE",
          "link": 107,
          "label": "图像"
        }
      ],
      "properties": {
        "Node name for S&R": "PreviewImage"
      }
    }
  ],
  "links": [
    [
      13,
      5,
      0,
      15,
      0,
      "LATENT"
    ],
    [
      14,
      12,
      0,
      17,
      1,
      "CONDITIONING"
    ],
    [
      16,
      13,
      0,
      18,
      0,
      "NOISE"
    ],
    [
      17,
      17,
      0,
      18,
      1,
      "GUIDER"
    ],
    [
      18,
      15,
      0,
      18,
      4,
      "LATENT"
    ],
    [
      19,
      14,
      0,
      18,
      2,
      "SAMPLER"
    ],
    [
      20,
      16,
      0,
      18,
      3,
      "SIGMAS"
    ],
    [
      22,
      10,
      0,
      19,
      1,
      "VAE"
    ],
    [
      34,
      37,
      0,
      12,
      1,
      "STRING"
    ],
    [
      35,
      37,
      0,
      12,
      2,
      "STRING"
    ],
    [
      73,
      8,
      0,
      11,
      0,
      "MODEL"
    ],
    [
      78,
      11,
      0,
      17,
      0,
      "MODEL"
    ],
    [
      79,
      11,
      0,
      16,
      0,
      "MODEL"
    ],
    [
      81,
      9,
      0,
      12,
      0,
      "CLIP"
    ],
    [
      89,
      19,
      0,
      7,
      0,
      "IMAGE"
    ],
    [
      93,
      18,
      0,
      19,
      0,
      "LATENT"
    ],
    [
      98,
      60,
      0,
      59,
      3,
      "SIGMAS"
    ],
    [
      99,
      11,
      0,
      61,
      0,
      "*"
    ],
    [
      100,
      61,
      0,
      60,
      0,
      "MODEL"
    ],
    [
      101,
      17,
      0,
      59,
      1,
      "GUIDER"
    ],
    [
      103,
      14,
      0,
      59,
      2,
      "SAMPLER"
    ],
    [
      104,
      10,
      0,
      63,
      0,
      "*"
    ],
    [
      105,
      63,
      0,
      62,
      1,
      "VAE"
    ],
    [
      106,
      59,
      0,
      62,
      0,
      "LATENT"
    ],
    [
      107,
      62,
      0,
      64,
      0,
      "IMAGE"
    ],
    [
      108,
      18,
      0,
      65,
      0,
      "LATENT"
    ],
    [
      109,
      65,
      0,
      59,
      4,
      "LATENT"
    ],
    [
      110,
      66,
      0,
      59,
      0,
      "NOISE"
    ]
  ],
  "groups": [],
  "config": {},
  "extra": {
    "ds": {
      "scale": 1.4864362802414455,
      "offset": [
        -818.3005016409714,
        -63.35849743821387
      ]
    }
  },
  "version": 0.4
}
CarolineSpike
关注
  • 13
    点赞
  • 踩
  • 4
    收藏
    觉得还不错? 一键收藏
  • 打赏
    打赏
  • 0
    评论
专栏目录
deep-learning-koans：使用Julia和Flux.jl进行的一系列教育性深度学习Koans
02-04
deep-learning-koans：使用Julia和Flux.jl进行的一系列教育性深度学习Koans
FlOp.jl:使用Optim.jl例程训练Flux.jl模型的函数
04-11
FlOp软件包包含一些功能，这些功能使人们可以使用带有或不带有GPU设施的Optim.jl中的优化例程来训练Flux.jl模型。 版本号 Juliav1.6.0 CUDA.jl v2.6.2 Flux.jl v0.12.0 Zygote.jl v0.6.7 Optim.jl v1.3.0 ...
FLUX.1最强AI绘画开源新模型，本地部署教程！
Yu91an的博客
08-05 1738
整体来说，测试下来效果还是很不错的，FLUX 比 SD3 强。手部生成比较稳定，较少出现手脚畸形。蒸馏版出图速度快， 30 秒内即可完成。Dev版虽然出图比较久但是出图质量更好。模型兼容性强：无论是二次元、人像、写实、还是风景风格，效果都很好。
FLUX.1 最强AI绘画开源模型，最新本地部署教程
RamendeusStudio的博客
08-06 942
Stable Diffusion 作者之一 Robin Rombach 官宣创业，成立了新公司「Black Forest Labs」（黑森林实验室），并推出了最新文生图模型「FLUX.1」。Black Forest Labs 在 AI 开发方面拥有丰富的经验，包括 VQGAN 和 Latent Diffusion 模型等开创性成果。该团队的经验涵盖学术、工业和开源环境，使其能够以独特的方式应对下一代 AI 技术的挑战。
消费级显卡，17G显存，玩转图像生成模型FLUX.1！ModelScope
stephen147的博客
08-06 777
从模型架构上看，FLUX.1和SD3有很多相似之处，都是基于FlowingMatching调度的模型，都通过引入T5来增强prompt的依从性。而比较显著的区别在于：flux模型引入了一种叫DoubleStreamBlock的结构，具体来说在前几层layer中，采用了txt和img embedding独立过各自的transformer块，然后再拼到一起过统一的transformer块，我们推测是为了进一步对齐图像和文本特征，但是这部分还没有更具体的技术报告。风格上，个人认为风格属性一般。
【实战】FLUX.1+ComfyUI部署与使用
最新发布
YourMi_AI的博客
08-15 533
FLUX.1是由黑森林实验室（根据官方资料：FLUX.1模型的训练参数达到了120亿，这一数字远远超过了SD3 Medium的20亿参数。无论是FLUX.1-pro还是FLUX.1-dev，在图像质量、提示词的准确跟随、尺寸适应性、排版以及输出的多样性方面，都已经超越了市场上的Midjourney v6.0、DALL·E 3 (HD)和SD3-Ultra等流行模型，确立了图像合成技术的新高度。
【AI辅助设计】全村人的希望！FLUX.1模型（附录下载地址）
2401_85688943的博客
08-02 1620
该项目是由原Stablity AI 公司的Robin 带着全新组织 Black Forest Labs，开源的一个AI生图模型。FLUX.1 模型套件为不同用户群体提供了灵活的选择，无论是追求极致性能的商业用户，还是希望探索 AI 技术的开发者和研究者，都能找到合适的版本。其强大的性能和丰富的功能，使其成为当前最先进的文本到图像模型之一，并将推动生成式 AI 技术的进一步发展。
FLUX.1 实测，堪比 Midjourney 的开源 AI 绘画模型，无需本地显卡，带你免费实战
u010522887的专栏
08-07 1755
玩转地表最强开源 AI 绘画模型
【震撼来袭】Flux横空出世，吊打Midjourney，重塑AI艺术创作新格局！（附保姆级部署教程）
Little_Carter的博客
08-05 820
近日，Midjourney 6.1的问世尚在业界掀起波澜，文生图开源领域的新贵——Flux 1.0便以迅雷不及掩耳之势，以其卓越的性能震撼登场。根据最新测试数据，Flux 1.0不仅在性能上大幅超越了DALL·E-3和Midjourney V6的闭源模型，更是将开源SD3系列的Ultra、Medium、Turbo和SDXL全线击溃，成为业界瞩目的焦点。Flux 1.0的问世，不仅仅是一个技术的飞跃，更是对现有文生图技术的一次革命性挑战。它以无与伦比的创造力和想象力，重新定义了艺术与科技的结合。
超越sd3！比肩Midjourney-v6？AI绘画大模型FLUX1.0详细评测与本地部署方法（附安装文件）
玩AI的小胡子-博客
08-06 992
​。
评分超越MJ和SD？揭秘FLUX.1 AI绘画模型，实例评测不容错过（附体验地址）
2401_85688943的博客
08-05 770
整体使用来说，1、出图的手部比较稳定，没有那么容易出现坏手、坏脚2、出图速度快，基本上30S能完成出图（我用的是云端4090显卡）3、模型的兼容性强：不管是二次元、人像模型、写实、电商、风景等不同风格的出图效果都可以如果说非要做个对比，我感觉Flux这个版本会比SD3强（目前SD3放出来的版本也不是质量最好的），但是说比上MJ，那还是有差距的。最近Mindjpurney V6.1的优化版本，出的图真的香。
flux.rar_flux_磁链_磁链模型_磁链观测
07-15
1. "flux.mdl"：这很可能是一个模型文件，可能是Flux软件使用的仿真或计算模型。MDL文件扩展名通常关联到各种不同的建模和设计软件，如Flux电磁学软件，它用于模拟电磁场的行为。在这个上下文中，该模型可能描述了一...
AlphaGo.jl：使用Flux.jl实现AlphaGo零
02-03
AlphaGo.jl是使用Flux.jl的AlphaGo Zero的纯Julia实现。 安装 要安装此软件包，只需运行 pkg> add https://github.com/tejank10/AlphaGo.jl 用法 using AlphaGo 创建Go环境很简单env = GoEnv(9) 这里9是木板的大小...
Julia-Flux-Demo:西蒙合奏团的Flux.jl演示
03-08
在这个Julia-Flux-Demo中，使用Jupyter Notebook可以方便地展示Flux.jl的代码和结果，使用户能够逐步理解模型的构建过程和训练效果。通过在Notebook中运行代码单元格，我们可以实时查看模型的性能指标、训练曲线以及...
Datawhale AI 夏令营 第四期 AIGC Task1
weixin_43680622的博客
08-10 979
Datawhale AI 夏令营（第四期）Task 1 从零入门AI生图原理&实践链接里的教程非常详细，很适合小白上手，从使用服务器平台到配置环境再到跑模型，手把手教！具体细节我就不赘述了，参看教程即可，下面我主要就此次Task1任务学到了哪些内容作一个总结。我开始接触图像生成呢，就是在进入2020年代后，大规模预训练模型时代。虽然是在这时候开始接触图像生成的，但并没有使用大模型，毕竟是小白嘛，就从经典的深度学习方法开始学习，如CNN和GAN。
Datawhale X 魔搭 AI夏令营第四期 | AIGC文生图——可图Kolors-LoRA风格故事 Task1笔记
ttrr27的博客
08-09 1126
Hi，大家好，我是半亩花海。最近在尝试学习AIGC的内容，并报名参加了Datawhale举办的2024年AI第四期夏令营，主要学习内容是从零入门AI生图原理和实践。本次活动基于魔搭社区“可图Kolors-LoRA风格故事挑战赛”进而开展的项目实践学习，很适合像我这样的想入门并实践AIGC文生图的初学者参与。
从互联网泡沫到AIGC：历史的教训与未来的机会
weixin_41496173的博客
08-13 714
在科技发展的长河中，泡沫现象是一个无法避免的阶段。1995年至2001年的互联网泡沫，是科技史上最具代表性的事件之一。而今天，随着生成式AI（AIGC）的迅猛崛起，许多人开始担忧其是否会重蹈互联网泡沫的覆辙。本文将通过回顾互联网泡沫的历史，分析AIGC与之相似的特征，探讨当前技术发展中的风险和机会。
Datawhale X 魔搭 AI夏令营第四期 AIGC方向 task02笔记
feishuoren的博客
08-14 464
AI工具使用
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

CarolineSpike

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值