Skip to main content
Render AI-generated videos on a dedicated cloud GPU. AnimateDiff turns text prompts into animated sequences. Both tools are VRAM-hungry — a cloud GPU avoids tying up your local machine.

​
GPU requirements

ToolMin VRAMRecommended GPU
AnimateDiff (SD 1.5)12 GBRTX 4090
AnimateDiff (SDXL)20 GBRTX 4090
CogVideoX24 GB+A100 80 GB

​
1. Deploy a GPU instance

runcrate instances create --name video-gen --gpu RTX4090 --template ubuntu-devbox
runcrate instances status video-gen

​
2. Install AnimateDiff

runcrate ssh video-gen -- "
  cd /workspace &&
  git clone https://github.com/guoyww/AnimateDiff.git &&
  cd AnimateDiff &&
  pip install -r requirements.txt
"
Download the motion module and base model: 
runcrate ssh video-gen -- "
  cd /workspace/AnimateDiff &&
  mkdir -p models/Motion_Module models/StableDiffusion &&
  huggingface-cli download guoyww/animatediff-motion-adapter-v1-5-3 \
    --local-dir models/Motion_Module &&
  huggingface-cli download runwayml/stable-diffusion-v1-5 \
    --local-dir models/StableDiffusion/stable-diffusion-v1-5
"

​
3. Generate a video

runcrate ssh video-gen -- "cd /workspace/AnimateDiff && python -m scripts.animate \
  --config configs/prompts/v3/v3-1-T2V.yaml \
  --pretrained_model_path models/StableDiffusion/stable-diffusion-v1-5 \
  --L 16 --W 512 --H 512"

​
4. Download rendered videos

runcrate cp video-gen:/workspace/AnimateDiff/output/ ./rendered-videos/

​
Monitor GPU during rendering

runcrate ssh video-gen -- nvidia-smi
runcrate ssh video-gen -- "ls -la /workspace/AnimateDiff/output/"

​
Tips

  • AnimateDiff at 512x512 with 16 frames fits on an RTX 4090. Higher resolutions need an A100.
  • Use --L 32 for longer animations. Memory scales linearly with frame count.
  • Convert frames to video with ffmpeg: ffmpeg -framerate 12 -i frame_%04d.png -c:v libx264 output.mp4.
  • Attach a volume to persist models across sessions.

​
Cleanup

runcrate instances delete video-gen