GitHubHelp

Man2Dev/wan-setup

Containerized Wan 2.1/2.2 video generation with Podman and NVIDIA GPU support on Fedora Linux

★ 0Forks 1DockerfileGitHub ↗Compare

Project website ↗

aifedoranvidiapodmantext-to-videovideo-generationwan

README

Wan Video Generation - Podman Setup

Containerized Wan 2.1 / Wan 2.2 video generation with NVIDIA GPU support on Fedora Linux.

Hardware Requirements

Model VRAM RAM Resolution
Wan 2.1 T2V-1.3B 8GB+ 16GB 480p
Wan 2.2 TI2V-5B 24GB+ 32GB 720p

Note: Most consumer GPUs should use Wan 2.1.

Prerequisites

# 1. Verify NVIDIA driver
nvidia-smi

# 2. Install podman and nvidia-container-toolkit
sudo dnf install -y podman
sudo dnf config-manager addrepo --from-repofile \
  https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo
sudo dnf install -y nvidia-container-toolkit

# 3. Generate CDI spec
sudo nvidia-ctk cdi generate --output=/etc/cdi/nvidia.yaml

# 4. Verify GPU access in container
podman run --rm --security-opt=label=disable --device=nvidia.com/gpu=all \
  nvidia/cuda:12.1.0-base-ubuntu22.04 nvidia-smi

Setup

git clone https://github.com/Man2Dev/wan-setup.git
cd wan-setup
mkdir -p models output

Wan 2.1 (Recommended for most GPUs)

Download & Build

# Download model (~5GB)
huggingface-cli download Wan-AI/Wan2.1-T2V-1.3B --local-dir ./models/Wan2.1-T2V-1.3B

# Build container
podman build -f Containerfile.wan21 -t wan21 .

Generate Video

podman run --rm \
  --security-opt=label=disable \
  --device=nvidia.com/gpu=all \
  -v ./models:/app/models:Z \
  -v ./output:/app/output:Z \
  wan21 \
  --task t2v-1.3B \
  --size "832*480" \
  --ckpt_dir /app/models/Wan2.1-T2V-1.3B \
  --offload_model True \
  --t5_cpu \
  --save_file /app/output/video.mp4 \
  --prompt "A serene lake at sunset with mountains in the background"

Supported Resolutions

  • 832*480 (landscape)
  • 480*832 (portrait)

Wan 2.2 (High-end GPUs only)

Download & Build

# Download model (~15-20GB)
huggingface-cli download Wan-AI/Wan2.2-TI2V-5B --local-dir ./models/Wan2.2-TI2V-5B

# Build container
podman build -t wan22 .

Generate Video

podman run --rm \
  --security-opt=label=disable \
  --device=nvidia.com/gpu=all \
  -v ./models:/app/models:Z \
  -v ./output:/app/output:Z \
  wan22 \
  --task ti2v-5B \
  --size "1280*704" \
  --ckpt_dir /app/models/Wan2.2-TI2V-5B \
  --offload_model True \
  --convert_model_dtype \
  --t5_cpu \
  --save_file /app/output/video.mp4 \
  --prompt "A serene lake at sunset with mountains in the background"

Image-to-Video

podman run --rm \
  --security-opt=label=disable \
  --device=nvidia.com/gpu=all \
  -v ./models:/app/models:Z \
  -v ./output:/app/output:Z \
  -v /path/to/image.jpg:/app/input.jpg:Z \
  wan22 \
  --task ti2v-5B \
  --size "1280*704" \
  --image /app/input.jpg \
  --ckpt_dir /app/models/Wan2.2-TI2V-5B \
  --offload_model True \
  --convert_model_dtype \
  --t5_cpu \
  --save_file /app/output/video.mp4 \
  --prompt "The scene comes to life"

Supported Resolutions

  • 1280*704 (landscape)
  • 704*1280 (portrait)

Common Options

Option Description
--size Video resolution
--frame_num Number of frames (must be 4n+1, default: 121)
--offload_model True Offload to CPU (saves VRAM)
--t5_cpu Run text encoder on CPU
--sample_steps Sampling steps (default: 50)
--base_seed Random seed (-1 for random)

Debugging

# Interactive shell
podman run -it --rm \
  --security-opt=label=disable \
  --device=nvidia.com/gpu=all \
  -v ./models:/app/models:Z \
  --entrypoint /bin/bash \
  wan22

Troubleshooting

Error Cause Fix
"Killed" Out of RAM Add swap: sudo fallocate -l 32G /swapfile && sudo mkswap /swapfile && sudo swapon /swapfile
CUDA OOM Out of VRAM Use --offload_model True --t5_cpu or switch to Wan 2.1
"could not select device driver" CDI not configured Run sudo nvidia-ctk cdi generate --output=/etc/cdi/nvidia.yaml

License

This repo: GPL-3.0-or-later
Wan models: Apache 2.0

Contributors

Man2Dev

Issues