ltx-2-19b-distilled-fp8 安装教程（ComfyUI 版）

一、适用类型

模型类型：LTX-2 19B 蒸馏版 FP8 视频生成模型（文本 / 图像生成视频 + 音频）
运行平台：Windows 10+/Linux（Ubuntu 20.04+），仅支持 NVIDIA 显卡
硬件要求：
- 最低：8GB 显存（RTX 3070/4060Ti 及以上）
- 推荐：12GB+ 显存（RTX 3090/4070Ti/4080/4090 等）
- 系统：CUDA 11.7+（推荐 12.1+）
- 存储：50GB+ 空闲空间（模型 + 依赖）
软件环境：Python 3.10–3.11、ComfyUI ≥1.7.0、Git

二、安装目录（核心）

1. ComfyUI 主目录（自定义）

Windows：D:\ComfyUI（建议非系统盘）
Linux：~/ComfyUI

2. 模型存放目录（必须）

主模型：ComfyUI/models/checkpoints/ltx-2-19b-distilled-fp8.safetensors
空间上采样器：ComfyUI/models/latent_upscale_models/ltx-2-spatial-upscaler-x2-1.0.safetensors
文本编码器：ComfyUI/models/text_encoders/gemma-3-12b-it/

3. 插件目录

LTXVideo 插件：ComfyUI/custom-nodes/ComfyUI-LTXVideo

三、详细安装步骤（3 种方案）

方案一：ComfyUI Manager 一键安装（推荐新手）

启动 ComfyUI，按 Ctrl+M 打开 Manager 面板
进入 Install Custom Nodes，搜索 LTXVideo
点击安装，等待完成后重启 ComfyUI
下载模型：
- 主模型：Hugging Face 下载 ltx-2-19b-distilled-fp8.safetensors，放入 models/checkpoints
- 辅助模型：下载空间上采样器、文本编码器，放入对应目录

方案二：手动 Git 安装（开发者）

进入 ComfyUI 自定义节点目录：

bash

运行

# Windows
cd D:\ComfyUI\custom-nodes
# Linux
cd ~/ComfyUI/custom-nodes

克隆插件仓库：

bash

运行

git clone https://gitcode.com/GitHub_Trending/ComfyUI-LTXVideo.git

安装依赖：

bash

运行

cd ComfyUI-LTXVideo
pip install -r requirements.txt

方案三：PyTorch 原生部署（代码调用）

克隆 LTX-2 仓库：

bash

运行

git clone https://github.com/Lightricks/LTX-2.git
cd LTX-2

创建虚拟环境并安装依赖：

bash

运行

python -m venv venv
# Windows
venv\Scripts\activate
# Linux
source venv/bin/activate
uv sync --frozen

下载模型：

bash

运行

huggingface-cli download Lightricks/LTX-2 --include "ltx-2-19b-distilled-fp8.safetensors" --local-dir models/

运行示例脚本：

bash

运行

python ltx-pipelines/example.py --model_path models/ltx-2-19b-distilled-fp8.safetensors

四、安装验证

启动 ComfyUI，在节点库中找到 LTXVideo 节点
加载示例工作流：custom-nodes/ComfyUI-LTXVideo/example_workflows/ltx-2-19b-distilled-fp8.json
输入提示词，点击生成，若成功输出视频则安装完成

五、常见问题

模型未找到：检查模型是否放入 models/checkpoints，文件名是否正确
显存不足：降低分辨率（540p）、帧数（24fps）、时长（4s）、步数（20）
依赖冲突：使用虚拟环境，重新安装依赖

ComfyUI 工作流，专用于 ltx-2-19b-distilled-fp8，

包含：

文生视频
图生视频
带音频生成
低显存优化（12GB 也能跑）

你只需要：

复制下面整段 JSON → 打开 ComfyUI → 直接粘贴到画布上 即可自动加载。

一、ltx-2-19b-distilled-fp8 完整版工作流（推荐）

json

{"last_node_id":40,"nodes":[{"id":1,"type":"LTXLoadModel","pos":[50,70],"size":[330,170],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[{"name":"MODEL","type":"MODEL","links":[2],"shape":3},{"name":"VAE","type":"VAE","links":[4],"shape":3}],"properties":{"Node name for S&R":"LTXLoadModel"},"widgets_values":["ltx-2-19b-distilled-fp8.safetensors","ltx_vae.safetensors"]},{"id":2,"type":"LTXLoadTextEncoder","pos":[420,70],"size":[330,130],"flags":{},"order":1,"mode":0,"inputs":[],"outputs":[{"name":"TEXT_ENCODER","type":"CLIP","links":[3],"shape":3}],"properties":{"Node name for S&R":"LTXLoadTextEncoder"},"widgets_values":["t5xxl_fp8_e4m3.safetensors"]},{"id":3,"type":"LTXTextEncode","pos":[790,70],"size":[330,140],"flags":{},"order":2,"mode":0,"inputs":[{"name":"text_encoder","type":"CLIP","link":3}],"outputs":[{"name":"CONDITIONING","type":"CONDITIONING","links":[5],"shape":3}],"properties":{"Node name for S&R":"LTXTextEncode"},"widgets_values":["cinematic,beautiful scenery,smooth motion,4K,highly detailed,realistic"]},{"id":4,"type":"LTXEmptyLatentVideo","pos":[50,270],"size":[250,100],"flags":{},"order":3,"mode":0,"inputs":[],"outputs":[{"name":"LATENT","type":"LATENT","links":[6],"shape":3}],"properties":{"Node name for S&R":"LTXEmptyLatentVideo"},"widgets_values":[704,400,16]},{"id":5,"type":"LTXSampler","pos":[420,270],"size":[330,260],"flags":{},"order":4,"mode":0,"inputs":[{"name":"model","type":"MODEL","link":2},{"name":"vae","type":"VAE","link":4},{"name":"positive","type":"CONDITIONING","link":5},{"name":"latent_video","type":"LATENT","link":6}],"outputs":[{"name":"LATENT","type":"LATENT","links":[7],"shape":3}],"properties":{"Node name for S&R":"LTXSampler"},"widgets_values":["euler","normal",12,5.5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,"fixed","enable"]},{"id":6,"type":"VTAEDecode","pos":[420,560],"size":[250,100],"flags":{},"order":5,"mode":0,"inputs":[{"name":"vae","type":"VAE","link":4},{"name":"samples","type":"LATENT","link":7}],"outputs":[{"name":"VIDEO","type":"IMAGE","links":[8],"shape":3}],"properties":{"Node name for S&R":"VTAEDecode"},"widgets_values":["default"]},{"id":9,"type":"LTXAudioLM","pos":[50,650],"size":[310,140],"flags":{},"order":6,"mode":0,"inputs":[],"outputs":[{"name":"AUDIO_LM","type":"AUDIO_MODEL","links":[10],"shape":3}],"properties":{"Node name for S&R":"LTXAudioLM"},"widgets_values":["lm_small_v1.safetensors"]},{"id":10,"type":"LTXVocoder","pos":[50,800],"size":[310,140],"flags":{},"order":7,"mode":0,"inputs":[],"outputs":[{"name":"VOCODER","type":"AUDIO_MODEL","links":[11],"shape":3}],"properties":{"Node name for S&R":"LTXVocoder"},"widgets_values":["vocoder_v1.safetensors"]},{"id":11,"type":"LTXGenerateAudio","pos":[420,670],"size":[310,140],"flags":{},"order":8,"mode":0,"inputs":[{"name":"AUDIO_LM","type":"AUDIO_MODEL","link":10},{"name":"VOCODER","type":"AUDIO_MODEL","link":11}],"outputs":[{"name":"AUDIO","type":"AUDIO","links":[12],"shape":3}],"properties":{"Node name for S&R":"LTXGenerateAudio"},"widgets_values":["cinematic ambient sound, calm background music",16]},{"id":12,"type":"PreviewVideoAudio","pos":[790,560],"size":[290,300],"flags":{},"order":9,"mode":0,"inputs":[{"name":"video","type":"IMAGE","link":8},{"name":"audio","type":"AUDIO","link":12}],"outputs":[],"properties":{"Node name for S&R":"PreviewVideoAudio"},"widgets_values":["ltx_distilled_output","video/mp4",1]}],"links":[[2,1,0,5,0],[3,2,0,3,0],[4,1,1,5,1],[5,3,0,5,2],[6,4,0,5,3],[7,5,0,6,1],[8,6,0,12,0],[10,9,0,11,0],[11,10,0,11,1],[12,11,0,12,1]],"groups":[],"config":{},"extra":{},"version":0.4}

二、图生视频专用工作流（蒸馏版）

json

{"last_node_id":42,"nodes":[{"id":1,"type":"LTXLoadModel","pos":[50,70],"size":[330,170],"flags":{},"order":0,"mode":0,"inputs":[],"outputs":[{"name":"MODEL","type":"MODEL","links":[2],"shape":3},{"name":"VAE","type":"VAE","links":[4],"shape":3}],"properties":{"Node name for S&R":"LTXLoadModel"},"widgets_values":["ltx-2-19b-distilled-fp8.safetensors","ltx_vae.safetensors"]},{"id":2,"type":"LTXLoadTextEncoder","pos":[420,70],"size":[330,130],"flags":{},"order":1,"mode":0,"inputs":[],"outputs":[{"name":"TEXT_ENCODER","type":"CLIP","links":[3],"shape":3}],"properties":{"Node name for S&R":"LTXLoadTextEncoder"},"widgets_values":["t5xxl_fp8_e4m3.safetensors"]},{"id":3,"type":"LTXTextEncode","pos":[790,70],"size":[330,140],"flags":{},"order":2,"mode":0,"inputs":[{"name":"text_encoder","type":"CLIP","link":3}],"outputs":[{"name":"CONDITIONING","type":"CONDITIONING","links":[5],"shape":3}],"properties":{"Node name for S&R":"LTXTextEncode"},"widgets_values":["make the image move naturally, smooth motion, cinematic"]},{"id":4,"type":"LoadImage","pos":[50,270],"size":[310,280],"flags":{},"order":3,"mode":0,"inputs":[],"outputs":[{"name":"IMAGE","type":"IMAGE","links":[6],"shape":3}],"properties":{"Node name for S&R":"LoadImage"},"widgets_values":["your_image.png","image"]},{"id":5,"type":"LTXImageToVideoLatent","pos":[420,270],"size":[330,120],"flags":{},"order":4,"mode":0,"inputs":[{"name":"IMAGE","type":"IMAGE","link":6}],"outputs":[{"name":"LATENT","type":"LATENT","links":[7],"shape":3}],"properties":{"Node name for S&R":"LTXImageToVideoLatent"},"widgets_values":[16]},{"id":6,"type":"LTXSampler","pos":[420,420],"size":[330,260],"flags":{},"order":5,"mode":0,"inputs":[{"name":"model","type":"MODEL","link":2},{"name":"vae","type":"VAE","link":4},{"name":"positive","type":"CONDITIONING","link":5},{"name":"latent_video","type":"LATENT","link":7}],"outputs":[{"name":"LATENT","type":"LATENT","links":[8],"shape":3}],"properties":{"Node name for S&R":"LTXSampler"},"widgets_values":["euler","normal",12,5.5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,"fixed","enable"]},{"id":7,"type":"VTAEDecode","pos":[420,710],"size":[250,100],"flags":{},"order":6,"mode":0,"inputs":[{"name":"vae","type":"VAE","link":4},{"name":"samples","type":"LATENT","link":8}],"outputs":[{"name":"VIDEO","type":"IMAGE","links":[9],"shape":3}],"properties":{"Node name for S&R":"VTAEDecode"},"widgets_values":["default"]},{"id":8,"type":"PreviewVideo","pos":[790,710],"size":[290,260],"flags":{},"order":7,"mode":0,"inputs":[{"name":"video","type":"IMAGE","link":9}],"outputs":[],"properties":{"Node name for S&R":"PreviewVideo"},"widgets_values":["ltx_img2vid_distilled","video/mp4",1]}],"links":[[2,1,0,6,0],[3,2,0,3,0],[4,1,1,6,1],[5,3,0,6,2],[6,4,0,5,0],[7,5,0,6,3],[8,6,0,7,1],[9,7,0,8,0]],"groups":[],"config":{},"extra":{},"version":0.4}

三、必须正确的模型目录（非常重要）

1. 主模型

plaintext

ComfyUI/models/ltx_video/ltx-2-19b-distilled-fp8.safetensors

2. VAE

plaintext

ComfyUI/models/vae/ltx_vae.safetensors

3. 文本编码器 T5

plaintext

ComfyUI/models/clip/t5xxl_fp8_e4m3.safetensors

4. 音频模型

plaintext

ComfyUI/models/audio_models/lm_small_v1.safetensors
ComfyUI/models/audio_models/vocoder_v1.safetensors

5. 插件位置

plaintext

ComfyUI/custom_nodes/ComfyUI-LTXVideo

四、蒸馏版最佳参数（直接照抄）

分辨率：704 × 400
帧数：16 帧
步数：12
CFG：5.5
采样器：euler
Offload：enable（必须开，省显存）

五、如果你粘贴后报错

一般就 2 个原因：

模型路径不对 / 文件名不对
没装 ComfyUI-LTXVideo 插件

#推荐
ltx-2-19b-distilled-fp8

ltx-2-19b-distilled-fp8 安装教程（ComfyUI 版）

一、适用类型

二、安装目录（核心）

1. ComfyUI 主目录（自定义）

2. 模型存放目录（必须）

3. 插件目录

三、详细安装步骤（3 种方案）

方案一：ComfyUI Manager 一键安装（推荐新手）

方案二：手动 Git 安装（开发者）

方案三：PyTorch 原生部署（代码调用）

四、安装验证

五、常见问题

ComfyUI 工作流，专用于 ltx-2-19b-distilled-fp8，

包含：

一、ltx-2-19b-distilled-fp8 完整版工作流（推荐）

二、图生视频专用工作流（蒸馏版）

三、必须正确的模型目录（非常重要）

1. 主模型

2. VAE

3. 文本编码器 T5

4. 音频模型

5. 插件位置

四、蒸馏版最佳参数（直接照抄）

五、如果你粘贴后报错

相关文章