FramePack

Next-Frame Prediction for Efficient Video Generation

Packing Input Frame Context in Next-Frame Prediction Models for Video Generation

View PaperGet Started

Try FramePack Today

Experience the future of video generation with minimal hardware requirements

FramePack - Image to Video

Drop your image here or click to browse

PNG, JPG or WEBP (max. 10MB)

Try FramePack Today - Hugging Face Demo

Experience the future of video generation with minimal hardware requirements

Key Features

⚡

Constant Workload

Compresses input contexts to a constant length so the generation workload is invariant to video length

💻

Low GPU Requirements

Process large numbers of frames with 13B models even on laptop GPUs with just 6GB memory

🔄

Large Batch Training

Can be trained with a much larger batch size, similar to image diffusion training

🎬

Progressive Generation

See results immediately as videos are generated frame-by-frame or section-by-section

System Requirements

Minimal standalone high-quality sampling system with efficient memory management

  • •

    Nvidia GPU in RTX 30XX, 40XX, 50XX series that supports fp16 and bf16

  • •

    Linux or Windows operating system

  • •

    At least 6GB GPU memory

Installation

Windows

  1. Download the One-Click Package (CUDA 12.6 + Pytorch 2.6)
  2. Uncompress the downloaded file
  3. Run update.bat to update to the latest version
  4. Use run.bat to start the application

Linux

  1. We recommend having an independent Python 3.10
  2. Install PyTorch with CUDA support
  3. Install requirements using pip install -r requirements.txt
  4. Start the GUI with python demo_gradio.py

Example Videos

FramePack can generate diverse, high-quality videos from a single image and text prompt

A car driving through a futuristic city at sunset, showcasing smooth next-frame video generation with light reflections and depth.

A car driving through a futuristic city at sunset, showcasing smooth next-frame video generation with light reflections and depth.

A three-frame sequence of a woman's face turning from front to side, illustrating detailed and consistent facial generation.

A three-frame sequence of a woman's face turning from front to side, illustrating detailed and consistent facial generation.

A seamless transition from sunset to night over a lake, highlighting realistic natural lighting and temporal consistency.

A seamless transition from sunset to night over a lake, highlighting realistic natural lighting and temporal consistency.

Try FramePack Today

Experience the future of video generation with minimal hardware requirements

Get Started