Progressive generation
FramePack builds video in successive frame sections, following the scene forward from your source image.
FramePack image-to-video
Turn a still image into a moving scene. Upload your starting frame, write a concise motion prompt, choose the output settings, and generate with FramePack online.
Output
Your video will appear here
Add an image and prompt, then choose your output settings.
The model, explained
FramePack is a next-frame prediction architecture for video diffusion. Instead of producing an entire clip in one pass, it generates the video progressively in frame sections.
Its frame-context packing design compresses the input context to a constant length. The official project describes this as keeping the generation workload stable as the requested sequence becomes longer.
This online interface connects the complete generation workflow in one place. You can upload an image, describe the motion, choose practical output settings, and follow the generation result from one workspace.
Built for image-to-video
A focused workflow for turning one strong frame into a controlled moving sequence.
FramePack builds video in successive frame sections, following the scene forward from your source image.
The architecture compresses prior context instead of allowing it to grow with every generated frame.
Short prompts centered on subject movement, scene action, and camera motion work naturally with the model.
Choose landscape or portrait, 480p or 720p, duration, seed, and guidance settings in one form.
The required credits update with duration and resolution, so you can review the cost before submitting.
Every request is sent with the provider's safety checker enabled; it cannot be disabled from the interface.
Three simple steps
Start with a clear image and describe only the motion you want to see.
Choose a sharp JPG, PNG, WebP, GIF, or AVIF image. A clear subject and readable composition give the model a stronger starting point.
Write one concise action: who or what moves, how it moves, and any important camera movement or atmosphere.
Select aspect ratio, resolution, and duration. Review the exact credit cost, submit, then preview and download the finished video.
The official project recommends concise, motion-focused prompts. These practical rules help keep the generated action clear.
Questions, answered
FramePack turns a starting image and a motion prompt into a video. Its next-frame prediction architecture extends the sequence progressively rather than generating every frame at once.
Your settings are validated by our server and submitted to a hosted FramePack endpoint. Credits are reserved when the job starts and returned automatically if submission or generation fails.
Generation time varies with duration, resolution, provider capacity, and queue length. The page shows whether your request is queued, processing, completed, or failed.
The cost depends on video duration and resolution. The generator recalculates and displays the exact number of credits before you submit the request.
Upload a frame, write one focused movement, and review the exact generation cost before you begin.
Create with FramePack