ComfyDeploy: How ComfyUI-Lumina-Video works in ComfyUI?
What is ComfyUI-Lumina-Video?
This is a video generation plugin implementation for ComfyUI based on the Lumina Video model.
How to install it in ComfyDeploy?
Head over to the machine page
- Click on the "Create a new machine" button
- Select the
Edit
build steps - Add a new step -> Custom Node
- Search for
ComfyUI-Lumina-Video
and select it - Close the build step dialig and then click on the "Save" button to rebuild the machine
ComfyUI-Lumina-Video
这是一个用于ComfyUI的基于 Lumina Video 模型的视频生成非官方插件实现。
功能特点
- 支持文本到视频的生成
- 支持自定义分辨率和帧率(宽高都必须为32的倍数)
- 支持自定义系统提示词和负面提示词
- 支持多种精度模式 (bf16/fp16/fp32)
- 自动下载并管理所需模型
安装方法
- 确保已安装 ComfyUI。(仅在linux 24GPU测试,估算16G可运行;欢迎issue反馈或pull request)
- 克隆本仓库到 ComfyUI 的 custom_nodes 目录:
cd ComfyUI/custom_nodes
git clone https://github.com/greengerong/ComfyUI-Lumina-Video.git
pip install -r requirements.txt
另外加速还需安装flash-attn
、Apex
,参加文档https://github.com/Alpha-VLLM/Lumina-Video/blob/main/INSTALL.md
- 重启ComfyUI,模型将在首次运行时自动下载。手动下载模型,则可用网盘下载并与本地models目录合并。网盘地址:https://pan.quark.cn/s/7663d2202040。
- Lumina-Video-f24R960 :下载模型并发至目录 /ComfyUI/models/Lumina-Video/Lumina-Video-f24R960
- gemma-2-2b :下载模型并发至目录 /ComfyUI/models/LLM/gemma-2-2b
- CogVideoX-2b vae :下载vae目录模型并发至目录 /ComfyUI/models/CogVideo/CogVideoX-2b
使用方法
工作流见./comfyui/workflows/Lumina-Video.json
<img width="1012" alt="demo" src="https://github.com/user-attachments/assets/79274471-77c2-4e13-bdd8-895c2f1a0b80" />