当前位置: 首页 > news >正文

5B 参数,消费级显卡可部署:Wan2.2-TI2V-5B 本地部署教程,9分钟跑出电影级大片!

一、模型介绍

Wan2.2-TI2V-5B 是阿里巴巴通义万相(Tongyi Wanxiang)团队于 2025 年 7 月开源的一款轻量级统一视频生成模型,属于 Wan2.2 系列中的核心成员。它以高效部署和多功能生成为特点,显著降低了电影级视频制作的技术门槛。

  • 核心定位与功能亮点

  1. 统一生成能力

    • 多模态输入支持:同时支持文本生成视频(Text-to-Video)和图文结合生成视频(Image+Text-to-Video),实现“二合一”功能,满足多样化的创作需求。

    • 电影级美学控制:用户可通过调整光影、色彩、构图等超 60 个参数,精细化控制画面风格,生成具有电影质感的动态内容。

  2. 轻量化设计

    • 参数规模仅 5B(50 亿参数),远小于同系列 MoE 架构模型(如 T2V-A14B 的 27B),但通过高压缩技术实现高性能输出。

  • 技术架构创新

  1. 高压缩 3D VAE 架构

    • 采用 时空压缩比 4×16×16 的 3D 变分自编码器,信息压缩率提升至 64 倍(远超前代 Wan2.1 的 4×8×8),显著减少显存占用。

    • 技术优势:在保持 720P 高清画质(24fps)的同时,大幅提升生成效率。

  2. 消费级硬件适配

    • 仅需 22GB 显存(如单张 RTX 4090)即可运行,支持共享显存模式(最低 8GB 显存可生成,但速度较慢)。

    • 生成耗时:单次生成 5 秒视频仅需数分钟,是目前开源模型中速度最快的 720P 视频生成方案之一。

  • 性能与效率表现

指标参数/性能对比优势
分辨率与帧率720P @ 24fps支持高清流畅输出
单次生成时长5 秒满足短视频、分镜需求
显存占用22GB(推荐) / 8GB(最低)消费级显卡可部署
生成速度≈2.5 分钟/20 步(RTX 4090)较同系列 A14B 模型快 50% 以上
  • Wan2.2系列核心型号及功能对比

型号Wan2.2-T2V-A14BWan2.2-I2V-A14BWan2.2-TI2V-5B
类型文生视频(Text-to-Video)图生视频(Image-to-Video)统一视频生成(Text/Image-to-Video)
参数量总参数27B,激活参数14B总参数27B,激活参数14B5B(密集架构)
核心架构MoE(混合专家)架构MoE架构高压缩3D VAE架构
生成质量电影级(光影/构图/微表情)电影级(细节优化突出)高清720P(流畅性优先)
美学控制✅ 支持60+参数精细调节✅ 支持同等美学控制⚠️ 部分支持(依赖提示词)
硬件要求高性能GPU(显存≥80GB)高性能GPU(显存≥80GB)消费级GPU(22GB显存)
生成速度5秒视频约15分钟(高性能卡)类似T2V-A14B5秒视频≤9分钟(RTX 4090)
开源平台GitHub/HuggingFace/魔搭社区同左同左

二、本地部署

本次部署采用 ComfyUI 作为本地部署框架。

环境版本号
Python3.12
PyTorch2.5.1
cuda12.1
Ubtuntu22.4.0

1.安装 Miniconda

步骤 1:更新系统
首先,更新系统软件包:

sudo apt update
sudo apt upgrade -y

步骤 2:下载 Miniconda 安装脚本
访问 Miniconda 的官方网站或使用以下命令直接下载最新版本的安装脚本(以 Python 3 为例):

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

步骤 3:验证安装脚本的完整性(可选)
下载 SHA256 校验和文件并验证安装包的完整性:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh.sha256
sha256sum Miniconda3-latest-Linux-x86_64.sh

比较输出的校验和与.sha256 文件中的值是否一致,确保文件未被篡改。

步骤 4:运行安装脚本
为安装脚本添加执行权限:

chmod +x Miniconda3-latest-Linux-x86_64.sh

运行安装脚本:

./Miniconda3-latest-Linux-x86_64.sh

步骤 5:按照提示完成安装
安装过程中需要:

阅读许可协议 :按 Enter 键逐页阅读,或者按 Q 退出阅读。
接受许可协议 :输入 yes 并按 Enter。
选择安装路径 :默认路径为/home/用户名/miniconda3,直接按 Enter 即可,或输入自定义路径。
是否初始化 Miniconda :输入 yes 将 Miniconda 添加到您的 PATH 环境变量中。
步骤 6:激活 Miniconda 环境
安装完成后,使环境变量生效:

source ~/.bashrc

步骤 7:验证安装是否成功
检查 conda 版本:

conda --version

步骤 8:更新 conda(推荐)
为了获得最新功能和修复,更新 conda:

conda update conda

2.部署 ComfyUI

2.1 克隆代码仓库

git clone https://github.com/comfyanonymous/ComfyUI.git

2.2 安装依赖

  • 创建 conda 虚拟环境
conda create -n comfyenv python==3.12
conda activate comfyenv
  • 安装 PyTorch
conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
  • 安装依赖
cd ComfyUI
pip install -r requirements.txt
  • 安装 ComfyUI Manager(可选)
#进入插件的文件
cd /ComfyUI/custom_nodes/
#下载ComfyUI Manager
git clone https://github.com/Comfy-Org/ComfyUI-Manager.git

3.下载 Wan2.2-TI2V-5B 模型相关文件

使用 modelscope 下载 Wan2.2-TI2V-5B 模型:以下是必要的模型文件

Diffusion Model wan2.2_ti2v_5B_fp16.safetensors

Text Encoder umt5_xxl_fp8_e4m3fn_scaled.safetensors

VAE wan2.2_vae.safetensors

需要将以上三个文件放入之前下载的 comfyUI 的对应目录下:

ComfyUI/
├───📂 models/
│ ├───📂 diffusion_models/
│ │ └───wan2.2_ti2v_5B_fp16.safetensors
│ ├───📂 text_encoders/
│ │ └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors
│ └───📂 vae/
│ └── wan2.2_vae.safetensors

对应下载命令:(local_dir 部分的路径可以自定义)

wan2.2_ti2v_5B_fp16.safetensors:

modelscope download --model Comfy-Org/Wan_2.2_ComfyUI_Repackaged --include '*/wan2.2_ti2v_5B_fp16.safetensors' --local_dir /models/diffusion_models/

umt5_xxl_fp8_e4m3fn_scaled.safetensors:

modelscope download --model Comfy-Org/Wan_2.2_ComfyUI_Repackaged --include '*/umt5_xxl_fp8_e4m3fn_scaled.safetensors' --local_dir /models/diffusion_models/

wan2.2_vae.safetensors:

modelscope download --model Comfy-Org/Wan_2.2_ComfyUI_Repackaged --include '*/wan2.2_vae.safetensors' --local_dir /models/diffusion_models/

4.启动 ComfyUI

之前下载的 comfyUI 根目录应该有 main.py 文件。

python main.py

运行 main.py 文件 之后根据控制台显示的 url 在浏览器输入该url网址进入 ComfyUI 的 GUI 界面:

To see the GUI go to: http://127.0.0.1:8188

如图:

1754880380998_6f7aee4c_14966762.png

5.使用 Wan2.2-TI2V-5B 工作流

依次点击 工作流 -> 打开,即可在电脑本地导入工作流 josn 文件。

1754880408073_c019f016_14966762.png

选择导入:

1754880419428_8ed828b9_14966762.png

完成,工作流文件在以下红色框住的位置点击后出现。然后选取想使用的 Wan2.2-TI2V-5B 工作流

1754880431834_ba9372fa_14966762.png

注: 当然也可以自主定制 工作流,后面放出三个工作流 json 文件内容作为参考。

Wan2.2-TI2V-5B 工作流文件内容

wan2.2_image_to_video.json
{"id": "91f6bbe2-ed41-4fd6-bac7-71d5b5864ecb","revision": 0,"last_node_id": 57,"last_link_id": 106,"nodes": [{"id": 8,"type": "VAEDecode","pos": [1210,190],"size": [210,46],"flags": {},"order": 10,"mode": 0,"inputs": [{"name": "samples","type": "LATENT","link": 35},{"name": "vae","type": "VAE","link": 76}],"outputs": [{"name": "IMAGE","type": "IMAGE","slot_index": 0,"links": [56,93]}],"properties": {"Node name for S&R": "VAEDecode"},"widgets_values": []},{"id": 7,"type": "CLIPTextEncode","pos": [413,389],"size": [425.27801513671875,180.6060791015625],"flags": {},"order": 6,"mode": 0,"inputs": [{"name": "clip","type": "CLIP","link": 75}],"outputs": [{"name": "CONDITIONING","type": "CONDITIONING","slot_index": 0,"links": [52]}],"title": "CLIP Text Encode (Negative Prompt)","properties": {"Node name for S&R": "CLIPTextEncode"},"widgets_values": ["色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,静止,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走"],"color": "#322","bgcolor": "#533"},{"id": 3,"type": "KSampler","pos": [863,187],"size": [315,262],"flags": {},"order": 9,"mode": 0,"inputs": [{"name": "model","type": "MODEL","link": 95},{"name": "positive","type": "CONDITIONING","link": 46},{"name": "negative","type": "CONDITIONING","link": 52},{"name": "latent_image","type": "LATENT","link": 104}],"outputs": [{"name": "LATENT","type": "LATENT","slot_index": 0,"links": [35]}],"properties": {"Node name for S&R": "KSampler"},"widgets_values": [869177064731501,"randomize",30,5,"uni_pc","simple",1]},{"id": 28,"type": "SaveAnimatedWEBP","pos": [1460,190],"size": [870.8511352539062,648.4141235351562],"flags": {},"order": 11,"mode": 0,"inputs": [{"name": "images","type": "IMAGE","link": 56}],"outputs": [],"properties": {},"widgets_values": ["ComfyUI",24.000000000000004,false,90,"default"]},{"id": 39,"type": "VAELoader","pos": [20,340],"size": [330,60],"flags": {},"order": 0,"mode": 0,"inputs": [],"outputs": [{"name": "VAE","type": "VAE","slot_index": 0,"links": [76,105]}],"properties": {"Node name for S&R": "VAELoader"},"widgets_values": ["wan2.2_vae.safetensors"],"color": "#223","bgcolor": "#335"},{"id": 38,"type": "CLIPLoader","pos": [20,190],"size": [380,106],"flags": {},"order": 1,"mode": 0,"inputs": [],"outputs": [{"name": "CLIP","type": "CLIP","slot_index": 0,"links": [74,75]}],"properties": {"Node name for S&R": "CLIPLoader"},"widgets_values": ["umt5_xxl_fp8_e4m3fn_scaled.safetensors","wan","default"],"color": "#223","bgcolor": "#335"},{"id": 48,"type": "ModelSamplingSD3","pos": [440,60],"size": [210,58],"flags": {},"order": 7,"mode": 0,"inputs": [{"name": "model","type": "MODEL","link": 94}],"outputs": [{"name": "MODEL","type": "MODEL","slot_index": 0,"links": [95]}],"properties": {"Node name for S&R": "ModelSamplingSD3"},"widgets_values": [8.000000000000002]},{"id": 37,"type": "UNETLoader","pos": [20,60],"size": [346.7470703125,82],"flags": {},"order": 2,"mode": 0,"inputs": [],"outputs": [{"name": "MODEL","type": "MODEL","slot_index": 0,"links": [94]}],"properties": {"Node name for S&R": "UNETLoader"},"widgets_values": ["wan2.2_ti2v_5B_fp16.safetensors","default"],"color": "#223","bgcolor": "#335"},{"id": 47,"type": "SaveWEBM","pos": [2367.213134765625,193.6114959716797],"size": [670,650],"flags": {},"order": 12,"mode": 0,"inputs": [{"name": "images","type": "IMAGE","link": 93}],"outputs": [],"properties": {"Node name for S&R": "SaveWEBM"},"widgets_values": ["ComfyUI","vp9",24,16.111083984375]},{"id": 57,"type": "LoadImage","pos": [87.407958984375,620.4816284179688],"size": [274.080078125,314],"flags": {},"order": 3,"mode": 0,"inputs": [],"outputs": [{"name": "IMAGE","type": "IMAGE","links": [106]},{"name": "MASK","type": "MASK","links": null}],"properties": {"Node name for S&R": "LoadImage"},"widgets_values": ["fennec_girl_hug.png","image"]},{"id": 56,"type": "Note","pos": [710.781005859375,608.9545288085938],"size": [320.9936218261719,182.6057586669922],"flags": {},"order": 4,"mode": 0,"inputs": [],"outputs": [],"properties": {},"widgets_values": ["Optimal resolution is: 1280x704 length 121\n\nThe reason it's lower in this workflow is just because I didn't want you to wait too long to get an initial video.\n\nTo get image to video just plug in a start image. For text to video just don't give it a start image."],"color": "#432","bgcolor": "#653"},{"id": 55,"type": "Wan22ImageToVideoLatent","pos": [420,610],"size": [271.9126892089844,150],"flags": {},"order": 8,"mode": 0,"inputs": [{"name": "vae","type": "VAE","link": 105},{"name": "start_image","shape": 7,"type": "IMAGE","link": 106}],"outputs": [{"name": "LATENT","type": "LATENT","links": [104]}],"properties": {"Node name for S&R": "Wan22ImageToVideoLatent"},"widgets_values": [1280,704,41,1]},{"id": 6,"type": "CLIPTextEncode","pos": [415,186],"size": [422.84503173828125,164.31304931640625],"flags": {},"order": 5,"mode": 0,"inputs": [{"name": "clip","type": "CLIP","link": 74}],"outputs": [{"name": "CONDITIONING","type": "CONDITIONING","slot_index": 0,"links": [46]}],"title": "CLIP Text Encode (Positive Prompt)","properties": {"Node name for S&R": "CLIPTextEncode"},"widgets_values": ["a cute anime girl with fennec ears and a fluffy tail walking in a beautiful field"],"color": "#232","bgcolor": "#353"}],"links": [[35,3,0,8,0,"LATENT"],[46,6,0,3,1,"CONDITIONING"],[52,7,0,3,2,"CONDITIONING"],[56,8,0,28,0,"IMAGE"],[74,38,0,6,0,"CLIP"],[75,38,0,7,0,"CLIP"],[76,39,0,8,1,"VAE"],[93,8,0,47,0,"IMAGE"],[94,37,0,48,0,"MODEL"],[95,48,0,3,0,"MODEL"],[104,55,0,3,3,"LATENT"],[105,39,0,55,0,"VAE"],[106,57,0,55,1,"IMAGE"]],"groups": [],"config": {},"extra": {"ds": {"scale": 1.1167815779425287,"offset": [3.5210927484772534,-9.231468990407302]},"frontendVersion": "1.23.4"},"version": 0.4
}
wan2.2_text_to_video.json
{"id": "91f6bbe2-ed41-4fd6-bac7-71d5b5864ecb","revision": 0,"last_node_id": 57,"last_link_id": 106,"nodes": [{"id": 8,"type": "VAEDecode","pos": [1210,190],"size": [210,46],"flags": {},"order": 9,"mode": 0,"inputs": [{"name": "samples","type": "LATENT","link": 35},{"name": "vae","type": "VAE","link": 76}],"outputs": [{"name": "IMAGE","type": "IMAGE","slot_index": 0,"links": [56,93]}],"properties": {"Node name for S&R": "VAEDecode"},"widgets_values": []},{"id": 7,"type": "CLIPTextEncode","pos": [413,389],"size": [425.27801513671875,180.6060791015625],"flags": {},"order": 6,"mode": 0,"inputs": [{"name": "clip","type": "CLIP","link": 75}],"outputs": [{"name": "CONDITIONING","type": "CONDITIONING","slot_index": 0,"links": [52]}],"title": "CLIP Text Encode (Negative Prompt)","properties": {"Node name for S&R": "CLIPTextEncode"},"widgets_values": ["色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,静止,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走"],"color": "#322","bgcolor": "#533"},{"id": 3,"type": "KSampler","pos": [863,187],"size": [315,262],"flags": {},"order": 8,"mode": 0,"inputs": [{"name": "model","type": "MODEL","link": 95},{"name": "positive","type": "CONDITIONING","link": 46},{"name": "negative","type": "CONDITIONING","link": 52},{"name": "latent_image","type": "LATENT","link": 104}],"outputs": [{"name": "LATENT","type": "LATENT","slot_index": 0,"links": [35]}],"properties": {"Node name for S&R": "KSampler"},"widgets_values": [285741127119524,"randomize",30,5,"uni_pc","simple",1]},{"id": 39,"type": "VAELoader","pos": [20,340],"size": [330,60],"flags": {},"order": 0,"mode": 0,"inputs": [],"outputs": [{"name": "VAE","type": "VAE","slot_index": 0,"links": [76,105]}],"properties": {"Node name for S&R": "VAELoader"},"widgets_values": ["wan2.2_vae.safetensors"],"color": "#223","bgcolor": "#335"},{"id": 38,"type": "CLIPLoader","pos": [20,190],"size": [380,106],"flags": {},"order": 1,"mode": 0,"inputs": [],"outputs": [{"name": "CLIP","type": "CLIP","slot_index": 0,"links": [74,75]}],"properties": {"Node name for S&R": "CLIPLoader"},"widgets_values": ["umt5_xxl_fp8_e4m3fn_scaled.safetensors","wan","default"],"color": "#223","bgcolor": "#335"},{"id": 48,"type": "ModelSamplingSD3","pos": [440,60],"size": [210,58],"flags": {},"order": 7,"mode": 0,"inputs": [{"name": "model","type": "MODEL","link": 94}],"outputs": [{"name": "MODEL","type": "MODEL","slot_index": 0,"links": [95]}],"properties": {"Node name for S&R": "ModelSamplingSD3"},"widgets_values": [8.000000000000002]},{"id": 37,"type": "UNETLoader","pos": [20,60],"size": [346.7470703125,82],"flags": {},"order": 2,"mode": 0,"inputs": [],"outputs": [{"name": "MODEL","type": "MODEL","slot_index": 0,"links": [94]}],"properties": {"Node name for S&R": "UNETLoader"},"widgets_values": ["wan2.2_ti2v_5B_fp16.safetensors","default"],"color": "#223","bgcolor": "#335"},{"id": 47,"type": "SaveWEBM","pos": [2367.213134765625,193.6114959716797],"size": [670,650],"flags": {},"order": 11,"mode": 0,"inputs": [{"name": "images","type": "IMAGE","link": 93}],"outputs": [],"properties": {"Node name for S&R": "SaveWEBM"},"widgets_values": ["ComfyUI","vp9",24,16.111083984375]},{"id": 56,"type": "Note","pos": [710.781005859375,608.9545288085938],"size": [320.9936218261719,182.6057586669922],"flags": {},"order": 3,"mode": 0,"inputs": [],"outputs": [],"properties": {},"widgets_values": ["Optimal resolution is: 1280x704 length 121\n\nThe reason it's lower in this workflow is just because I didn't want you to wait too long to get an initial video.\n\nTo get image to video just plug in a start image. For text to video just don't give it a start image."],"color": "#432","bgcolor": "#653"},{"id": 55,"type": "Wan22ImageToVideoLatent","pos": [420,610],"size": [271.9126892089844,150],"flags": {},"order": 4,"mode": 0,"inputs": [{"name": "vae","type": "VAE","link": 105},{"name": "start_image","shape": 7,"type": "IMAGE","link": null}],"outputs": [{"name": "LATENT","type": "LATENT","links": [104]}],"properties": {"Node name for S&R": "Wan22ImageToVideoLatent"},"widgets_values": [1280,704,41,1]},{"id": 6,"type": "CLIPTextEncode","pos": [415,186],"size": [422.84503173828125,164.31304931640625],"flags": {},"order": 5,"mode": 0,"inputs": [{"name": "clip","type": "CLIP","link": 74}],"outputs": [{"name": "CONDITIONING","type": "CONDITIONING","slot_index": 0,"links": [46]}],"title": "CLIP Text Encode (Positive Prompt)","properties": {"Node name for S&R": "CLIPTextEncode"},"widgets_values": ["drone shot of a volcano erupting with a fox walking on it"],"color": "#232","bgcolor": "#353"},{"id": 28,"type": "SaveAnimatedWEBP","pos": [1460,190],"size": [870.8511352539062,648.4141235351562],"flags": {},"order": 10,"mode": 0,"inputs": [{"name": "images","type": "IMAGE","link": 56}],"outputs": [],"properties": {},"widgets_values": ["ComfyUI",24.000000000000004,false,80,"default"]}],"links": [[35,3,0,8,0,"LATENT"],[46,6,0,3,1,"CONDITIONING"],[52,7,0,3,2,"CONDITIONING"],[56,8,0,28,0,"IMAGE"],[74,38,0,6,0,"CLIP"],[75,38,0,7,0,"CLIP"],[76,39,0,8,1,"VAE"],[93,8,0,47,0,"IMAGE"],[94,37,0,48,0,"MODEL"],[95,48,0,3,0,"MODEL"],[104,55,0,3,3,"LATENT"],[105,39,0,55,0,"VAE"]],"groups": [],"config": {},"extra": {"ds": {"scale": 1.11678157794253,"offset": [7.041966347099882,-19.733042401058505]},"frontendVersion": "1.23.4"},"version": 0.4
}
wan2.2_workflow.json
{"id": "91f6bbe2-ed41-4fd6-bac7-71d5b5864ecb","revision": 0,"last_node_id": 59,"last_link_id": 108,"nodes": [{"id": 37,"type": "UNETLoader","pos": [-30,50],"size": [346.7470703125,82],"flags": {},"order": 0,"mode": 0,"inputs": [],"outputs": [{"name": "MODEL","type": "MODEL","slot_index": 0,"links": [94]}],"properties": {"Node name for S&R": "UNETLoader","cnr_id": "comfy-core","ver": "0.3.45","models": [{"name": "wan2.2_ti2v_5B_fp16.safetensors","url": "https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors","directory": "diffusion_models"}]},"widgets_values": ["wan2.2_ti2v_5B_fp16.safetensors","default"]},{"id": 38,"type": "CLIPLoader","pos": [-30,190],"size": [350,110],"flags": {},"order": 1,"mode": 0,"inputs": [],"outputs": [{"name": "CLIP","type": "CLIP","slot_index": 0,"links": [74,75]}],"properties": {"Node name for S&R": "CLIPLoader","cnr_id": "comfy-core","ver": "0.3.45","models": [{"name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors","url": "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors","directory": "text_encoders"}]},"widgets_values": ["umt5_xxl_fp8_e4m3fn_scaled.safetensors","wan","default"]},{"id": 39,"type": "VAELoader","pos": [-30,350],"size": [350,60],"flags": {},"order": 2,"mode": 0,"inputs": [],"outputs": [{"name": "VAE","type": "VAE","slot_index": 0,"links": [76,105]}],"properties": {"Node name for S&R": "VAELoader","cnr_id": "comfy-core","ver": "0.3.45","models": [{"name": "wan2.2_vae.safetensors","url": "https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors","directory": "vae"}]},"widgets_values": ["wan2.2_vae.safetensors"]},{"id": 8,"type": "VAEDecode","pos": [1190,150],"size": [210,46],"flags": {},"order": 10,"mode": 0,"inputs": [{"name": "samples","type": "LATENT","link": 35},{"name": "vae","type": "VAE","link": 76}],"outputs": [{"name": "IMAGE","type": "IMAGE","slot_index": 0,"links": [107]}],"properties": {"Node name for S&R": "VAEDecode","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": []},{"id": 57,"type": "CreateVideo","pos": [1200,240],"size": [270,78],"flags": {},"order": 11,"mode": 0,"inputs": [{"name": "images","type": "IMAGE","link": 107},{"name": "audio","shape": 7,"type": "AUDIO","link": null}],"outputs": [{"name": "VIDEO","type": "VIDEO","links": [108]}],"properties": {"Node name for S&R": "CreateVideo","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": [24]},{"id": 58,"type": "SaveVideo","pos": [1200,370],"size": [660,450],"flags": {},"order": 12,"mode": 0,"inputs": [{"name": "video","type": "VIDEO","link": 108}],"outputs": [],"properties": {"Node name for S&R": "SaveVideo","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": ["video/ComfyUI","auto","auto"]},{"id": 55,"type": "Wan22ImageToVideoLatent","pos": [380,540],"size": [271.9126892089844,150],"flags": {},"order": 8,"mode": 0,"inputs": [{"name": "vae","type": "VAE","link": 105},{"name": "start_image","shape": 7,"type": "IMAGE","link": 106}],"outputs": [{"name": "LATENT","type": "LATENT","links": [104]}],"properties": {"Node name for S&R": "Wan22ImageToVideoLatent","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": [1280,704,121,1]},{"id": 56,"type": "LoadImage","pos": [0,540],"size": [274.080078125,314],"flags": {},"order": 3,"mode": 4,"inputs": [],"outputs": [{"name": "IMAGE","type": "IMAGE","links": [106]},{"name": "MASK","type": "MASK","links": null}],"properties": {"Node name for S&R": "LoadImage","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": ["example.png","image"]},{"id": 7,"type": "CLIPTextEncode","pos": [380,260],"size": [425.27801513671875,180.6060791015625],"flags": {},"order": 7,"mode": 0,"inputs": [{"name": "clip","type": "CLIP","link": 75}],"outputs": [{"name": "CONDITIONING","type": "CONDITIONING","slot_index": 0,"links": [52]}],"title": "CLIP Text Encode (Negative Prompt)","properties": {"Node name for S&R": "CLIPTextEncode","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": ["色调艳丽,过曝,静态,细节模糊不清,字幕,风格,作品,画作,画面,静止,整体发灰,最差质量,低质量,JPEG压缩残留,丑陋的,残缺的,多余的手指,画得不好的手部,画得不好的脸部,畸形的,毁容的,形态畸形的肢体,手指融合,静止不动的画面,杂乱的背景,三条腿,背景人很多,倒着走"],"color": "#322","bgcolor": "#533"},{"id": 59,"type": "MarkdownNote","pos": [-550,10],"size": [480,340],"flags": {},"order": 4,"mode": 0,"inputs": [],"outputs": [],"title": "Model Links","properties": {},"widgets_values": ["[Tutorial](https://docs.comfy.org/tutorials/video/wan/wan2_2\n) | [教程](https://docs.comfy.org/zh-CN/tutorials/video/wan/wan2_2\n)\n\n**Diffusion Model**\n- [wan2.2_ti2v_5B_fp16.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/diffusion_models/wan2.2_ti2v_5B_fp16.safetensors)\n\n**VAE**\n- [wan2.2_vae.safetensors](https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan2.2_vae.safetensors)\n\n**Text Encoder**   \n- [umt5_xxl_fp8_e4m3fn_scaled.safetensors](https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors)\n\n\nFile save location\n\n```\nComfyUI/\n├───📂 models/\n│   ├───📂 diffusion_models/\n│   │   └───wan2.2_ti2v_5B_fp16.safetensors\n│   ├───📂 text_encoders/\n│   │   └─── umt5_xxl_fp8_e4m3fn_scaled.safetensors \n│   └───📂 vae/\n│       └── wan2.2_vae.safetensors\n```\n"],"color": "#432","bgcolor": "#653"},{"id": 6,"type": "CLIPTextEncode","pos": [380,50],"size": [422.84503173828125,164.31304931640625],"flags": {},"order": 6,"mode": 0,"inputs": [{"name": "clip","type": "CLIP","link": 74}],"outputs": [{"name": "CONDITIONING","type": "CONDITIONING","slot_index": 0,"links": [46]}],"title": "CLIP Text Encode (Positive Prompt)","properties": {"Node name for S&R": "CLIPTextEncode","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": ["Low contrast. In a retro 1970s-style subway station, a street musician plays in dim colors and rough textures. He wears an old jacket, playing guitar with focus. Commuters hurry by, and a small crowd gathers to listen. The camera slowly moves right, capturing the blend of music and city noise, with old subway signs and mottled walls in the background."],"color": "#232","bgcolor": "#353"},{"id": 3,"type": "KSampler","pos": [850,130],"size": [315,262],"flags": {},"order": 9,"mode": 0,"inputs": [{"name": "model","type": "MODEL","link": 95},{"name": "positive","type": "CONDITIONING","link": 46},{"name": "negative","type": "CONDITIONING","link": 52},{"name": "latent_image","type": "LATENT","link": 104}],"outputs": [{"name": "LATENT","type": "LATENT","slot_index": 0,"links": [35]}],"properties": {"Node name for S&R": "KSampler","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": [898471028164125,"randomize",20,5,"uni_pc","simple",1]},{"id": 48,"type": "ModelSamplingSD3","pos": [850,20],"size": [210,58],"flags": {"collapsed": false},"order": 5,"mode": 0,"inputs": [{"name": "model","type": "MODEL","link": 94}],"outputs": [{"name": "MODEL","type": "MODEL","slot_index": 0,"links": [95]}],"properties": {"Node name for S&R": "ModelSamplingSD3","cnr_id": "comfy-core","ver": "0.3.45"},"widgets_values": [8]}],"links": [[35,3,0,8,0,"LATENT"],[46,6,0,3,1,"CONDITIONING"],[52,7,0,3,2,"CONDITIONING"],[74,38,0,6,0,"CLIP"],[75,38,0,7,0,"CLIP"],[76,39,0,8,1,"VAE"],[94,37,0,48,0,"MODEL"],[95,48,0,3,0,"MODEL"],[104,55,0,3,3,"LATENT"],[105,39,0,55,0,"VAE"],[106,56,0,55,1,"IMAGE"],[107,8,0,57,0,"IMAGE"],[108,57,0,58,0,"VIDEO"]],"groups": [{"id": 1,"title": "Step1 - Load models","bounding": [-50,-20,400,453.6000061035156],"color": "#3f789e","font_size": 24,"flags": {}},{"id": 2,"title": "Step3 - Prompt","bounding": [370,-20,448.27801513671875,473.2060852050781],"color": "#3f789e","font_size": 24,"flags": {}},{"id": 3,"title": "For i2v, use Ctrl + B to enable","bounding": [-50,450,400,420],"color": "#3f789e","font_size": 24,"flags": {}},{"id": 4,"title": "Video Size & length","bounding": [370,470,291.9127197265625,233.60000610351562],"color": "#3f789e","font_size": 24,"flags": {}}],"config": {},"extra": {"ds": {"scale": 0.8390545288824454,"offset": [252.56402822629002,99.17599442262053]},"frontendVersion": "1.25.1","VHS_latentpreview": false,"VHS_latentpreviewrate": 0,"VHS_MetadataImage": true,"VHS_KeepIntermediate": true},"version": 0.4
}
http://www.lryc.cn/news/617878.html

相关文章:

  • 【运维进阶】WEB 服务器
  • GB17761-2024标准与电动自行车防火安全的技术革新
  • Redis 数据结构及特点
  • 【工作笔记】Wrappers.lambdaQuery()用法
  • ROS2学习(1)—基础概念及环境搭建
  • JavaEE 初阶第十七期:文件 IO 的 “管道艺术”(下)
  • vscode新建esp32工程,没有sample_project怎么办?
  • 计算机网络:ovn数据通信都是用了哪些协议?
  • 应用层模拟面试题
  • C语言(06)——二、八、十、十六进制的相互转换
  • Effective C++ 条款35:考虑 virtual函数以外的其他选择
  • 【已解决】报错:WARNING: pip is configured with locations that require TLS/SSL
  • HarmonyOS 开发入门 第一章
  • 一文读懂 C# 中的 Lazy<T>
  • Python 在自动化办公汇总和脚本示例
  • 本地文件夹与 GitHub 远程仓库绑定并进行日常操作的完整命令流程
  • 【基本有序数组中找到有且仅有的一个无序元素并进行排序之顺序法】2022-10-12
  • Linux线程——线程控制及理解
  • Transformer前传:Seq2Seq与注意力机制Attention
  • Haystack:面向大模型应用的模块化检索增强生成(RAG)框架
  • 什么情况下会导致日本服务器变慢?解决办法
  • Linux kernel network stack, some good article
  • Flink + Hologres构建实时数仓
  • Spring JDBC
  • TDengine IDMP 基本功能(1.界面布局和操作)
  • 【华为机试】208. 实现 Trie (前缀树)
  • openGauss逻辑备份恢复工具gs_dump/gs_restore
  • AI生成代码时代的商业模式重构:从“软件即产品”到“价值即服务”
  • 大模型落地实践:从技术重构到行业变革的双重突破
  • 亚马逊广告底层逻辑重构:从流量博弈到价值创造的战略升维