FramePack
From SD's author lllyasviel — 1-minute videos on 6GB VRAM, a blessing for consumer GPUs
★ 17,266 stars, verified 2026-09-30Category: Video / Digital Humans / RestorationGitHub repository ↗
From Stable Diffusion author lllyasviel: generate 1-minute videos locally on 6GB of VRAM — a blessing for consumer GPUs.
AI video generation, digital humans, and video restoration.
Key features
- Next-frame-section prediction video diffusion model
- Compresses input context to a constant length, workload invariant to video length
- Runs 13B models on laptop GPUs with as little as 6GB VRAM
- One-click Windows package; Linux install via pip with Gradio GUI
- Optional sage-attention, xformers, flash-attn backends
Good for
- Generating long image-to-video clips on consumer GPUs
- Researchers studying next-frame video diffusion
Stack & tags
PythonPyTorchGradiovideo-generationdiffusionopen-weightsimage-to-videolow-vramgradio
Quick start
pip install -r requirements.txt
python demo_gradio.pyFrom the project README on GitHub.
Similar open-source projects
Hyperframes
★ 55,468Write HTML. Render video. Built for agents — HeyGen's open-source video framework
Hypit
★ 18,649Clone any viral video with AI agents — swap the face, the words, the B-roll, ship 100 variants