下载 ComfyUI Portable
下载地址： https://github.com/Comfyanonymous/ComfyUI/releases
选择适合你显卡的最新版本
解压位置： 任意目录，建议 D:\ComfyUI_windows_portable\
目录结构：
1
2
3
4
5
6
7
8
| ComfyUI_windows_portable/
├── python_embeded/ # 内置 Python 环境
├── ComfyUI/
│ ├── models/ # 模型存放目录
│ ├── custom_nodes/ # 插件目录
│ └── main.py
├── run_nvidia_gpu.bat # NVIDIA GPU 启动脚本
└── run_cpu.bat # CPU 启动脚本
|
安装必要插件
2.1 打开终端进入 ComfyUI 目录
1
| cd D:\ComfyUI_windows_portable\ComfyUI
|
2.2 安装 ComfyUI-GGUF 插件
1
2
3
4
5
6
| cd custom_nodes
git clone https://github.com/city96/ComfyUI-GGUF.git
cd ComfyUI-GGUF
# 使用内置 Python 安装依赖
..\..\python_embeded\python.exe -m pip install -r requirements.txt
|
2.3 安装 ComfyUI-KJNodes 插件
1
2
3
4
| cd ..
git clone https://github.com/kijai/ComfyUI-KJNodes.git
cd ComfyUI-KJNodes
..\..\python_embeded\python.exe -m pip install -r requirements.txt
|
2.4 安装 huggingface_hub
1
| ..\..\python_embeded\python.exe -m pip install huggingface_hub
|
下载模型文件
3.1 模型下载脚本
创建下载脚本 download_ltx_gguf.ps1，放在 ComfyUI_windows_portable\ComfyUI\ 目录：
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
| # download_ltx_gguf.ps1
# 使用方法：在 ComfyUI 目录运行 .\download_ltx_gguf.ps1
$modelsDir = ".\models"
$pythonExe = "..\python_embeded\python.exe"
Write-Host "开始下载 LTX-2.3-GGUF 模型文件..." -ForegroundColor Cyan
# 创建必要目录
$dirs = @(
"$modelsDir\unet",
"$modelsDir\vae",
"$modelsDir\text_encoders",
"$modelsDir\loras",
"$modelsDir\latent_upscale_models"
)
foreach ($dir in $dirs) {
if (!(Test-Path $dir)) {
New-Item -ItemType Directory -Path $dir -Force | Out-Null
Write-Host "创建目录: $dir" -ForegroundColor Green
}
}
Write-Host "`n下载 GGUF 主模型 (约 12GB)..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download unsloth/LTX-2.3-GGUF ltx-2.3-22b-dev-Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
Move-Item "ltx-2.3-22b-dev-Q4_K_M.gguf" "$modelsDir\unet\" -Force
Write-Host "`n下载视频 VAE..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download unsloth/LTX-2.3-GGUF vae/ltx-2.3-22b-dev_video_vae.safetensors --local-dir . --local-dir-use-symlinks False
Move-Item "vae\ltx-2.3-22b-dev_video_vae.safetensors" "$modelsDir\vae\" -Force
Write-Host "`n下载音频 VAE..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download unsloth/LTX-2.3-GGUF vae/ltx-2.3-22b-dev_audio_vae.safetensors --local-dir . --local-dir-use-symlinks False
Move-Item "vae\ltx-2.3-22b-dev_audio_vae.safetensors" "$modelsDir\vae\" -Force
Write-Host "`n下载 embeddings connectors..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download unsloth/LTX-2.3-GGUF text_encoders/ltx-2.3-22b-dev_embeddings_connectors.safetensors --local-dir . --local-dir-use-symlinks False
Move-Item "text_encoders\ltx-2.3-22b-dev_embeddings_connectors.safetensors" "$modelsDir\text_encoders\" -Force
Write-Host "`n下载 Gemma 文本编码器 GGUF..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download unsloth/gemma-3-12b-it-qat-GGUF gemma-3-12b-it-qat-UD-Q4_K_XL.gguf --local-dir . --local-dir-use-symlinks False
Move-Item "gemma-3-12b-it-qat-UD-Q4_K_XL.gguf" "$modelsDir\text_encoders\" -Force
& $pythonExe -m huggingface_hub.commands.huggingface_cli download unsloth/gemma-3-12b-it-qat-GGUF mmproj-BF16.gguf --local-dir . --local-dir-use-symlinks False
Move-Item "mmproj-BF16.gguf" "$modelsDir\text_encoders\" -Force
Write-Host "`n下载 Distilled LoRA..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download Lightricks/LTX-2.3 ltx-2.3-22b-distilled-lora-384.safetensors --local-dir . --local-dir-use-symlinks False
Move-Item "ltx-2.3-22b-distilled-lora-384.safetensors" "$modelsDir\loras\" -Force
Write-Host "`n下载空间放大器..." -ForegroundColor Yellow
& $pythonExe -m huggingface_hub.commands.huggingface_cli download Lightricks/LTX-2.3 ltx-2.3-spatial-upscaler-x2-1.0.safetensors --local-dir . --local-dir-use-symlinks False
Move-Item "ltx-2.3-spatial-upscaler-x2-1.0.safetensors" "$modelsDir\latent_upscale_models\" -Force
Write-Host "`n========================================" -ForegroundColor Cyan
Write-Host "所有模型下载完成！" -ForegroundColor Green
Write-Host "========================================" -ForegroundColor Cyan
|
3.2 运行下载脚本
1
2
3
| cd D:\ComfyUI_windows_portable\ComfyUI
Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\download_ltx_gguf.ps1
|
模型文件清单
下载完成后，目录结构应为：
1
2
3
4
5
6
7
8
9
10
11
12
13
14
| ComfyUI_windows_portable\ComfyUI\models\
├── unet\
│ └── ltx-2.3-22b-dev-Q4_K_M.gguf # ~12GB
├── vae\
│ ├── ltx-2.3-22b-dev_video_vae.safetensors
│ └── ltx-2.3-22b-dev_audio_vae.safetensors
├── text_encoders\
│ ├── ltx-2.3-22b-dev_embeddings_connectors.safetensors
│ ├── gemma-3-12b-it-qat-UD-Q4_K_XL.gguf # ~7GB
│ └── mmproj-BF16.gguf
├── loras\
│ └── ltx-2.3-22b-distilled-lora-384.safetensors
└── latent_upscale_models\
└── ltx-2.3-spatial-upscaler-x2-1.0.safetensors
|
总下载大小： 约 25-30GB
启动 ComfyUI
1
2
| cd D:\ComfyUI_windows_portable
.\run_nvidia_gpu.bat
|
启动后访问：http://127.0.0.1:8188
验证安装
6.1 检查插件是否加载
在 ComfyUI 界面中：
- 双击空白处打开节点搜索
- 搜索
GGUF - 应能看到
Unet Loader (GGUF) 等节点
6.2 检查模型是否识别
- 添加
Unet Loader (GGUF) 节点 - 点击模型下拉菜单
- 应能看到
ltx-2.3-22b-dev-Q4_K_M.gguf
