🎞️ FramePack Runner – Image-to-Video with Prompt

Generate short videos from a single image and a motion prompt using the FramePack architecture by Stanford (2025).

👉 Try it now on Replicate

✨ Example usage

replicate run paullux/framepack-runner \
  -v image=@input.png \
  -v prompt="A cat jumps backward in surprise" \
  -v seed=123 \
  -v steps=30 \
  -v duration_seconds=5 \
  -v fps=30

🧠 About FramePack

Packing Input Frame Contexts in Next-Frame Prediction Models for Video Generation Lvmin Zhang, Maneesh Agrawala – Stanford University, 2025

FramePack compresses temporal context into fixed-length representations, making it highly efficient for generating long video sequences.

📄 Project page
🧪 Uses the FramePackPipeline from 🤗 diffusers

📦 Inputs

Name	Type	Description
`image`	`file`	The input image (`.png` or `.jpg`)
`prompt`	`string`	A motion-focused description prompt
`seed`	`integer`	Random seed for reproducibility (default: `42`)
`steps`	`integer`	Sampling steps (default: `25`)
`duration_seconds`	`number`	Duration of the video in seconds (default: `5`)
`fps`	`integer`	Output video frame rate (default: `30`)

📤 Output

Returns an .mp4 video file composed of all generated frames.

🛠 How it works

python framepack_runner.py \
  -v image=@input.png \
  -v prompt="A cat jumps backward in surprise" \
  -v seed=123 \
  -v steps=30 \
  -v duration_seconds=5 \
  -v fps=30

All frames are saved locally and then compiled into a video using ffmpeg.

📸 Example Prompt Ideas

The robot jumps forward and transforms mid-air.
A woman spins slowly in a dark room, lit by candlelight.
The camera zooms toward a cat staring at a moving shadow.

📃 License

This implementation uses FramePack under its original academic license. Your usage of Replicate infrastructure is bound by their terms of service.

🙌 Credits

Based on FramePack by Lvmin Zhang & Maneesh Agrawala (Stanford).