FramePack Video Diffusion

Discover FramePack Video Diffusion, a revolutionary model for progressive video generation. Discover how this AI model can transform your workflow!

Platform: Replicate

Video DiffusionProgressive GenerationImage-to-Video

1.8k runs

A100 (80GB)

Commercial

🚀Function Overview

A model that generates videos progressively from an initial image using text prompts, implementing non-F1 FramePack video diffusion with efficient memory management.

Key Features

Progressive video generation in sections
Starts from input images with text-guided motion control
Efficient GPU memory management for various hardware
Smooth blending of video sections
Configurable quality vs. speed parameters

Use Cases

•Animating still images with text prompts
•Creating short dynamic video clips of motions/actions
•Visualizing scene evolutions like weather changes
•Prototyping story concepts and visual narratives

⚙️Input Parameters

input_image

string

Initial image to start the video from

prompt

string

Text description of what you want in the video

negative_prompt

string

Things you want to avoid in the video

seed

integer

Set for consistent results or leave empty for random

total_video_length_seconds

number

How long the video should be in seconds

latent_window_size

integer

Controls how video chunks are processed - smaller is faster but may reduce quality

steps

integer

More steps give higher quality but take longer

cfg_scale

number

Higher values follow the prompt more closely, lower values are more creative

distilled_cfg_scale

number

Controls prompt adherence for distilled model components

cfg_rescale

number

Reduces oversaturation at high CFG values

use_teacache

boolean

Makes generation faster with minimal quality impact

mp4_crf

integer

Controls video compression - lower values give better quality but larger files

💡Usage Examples

Example 1

Input Parameters

{
  "steps": 25,
  "prompt": "Breakdancing",
  "mp4_crf": 23,
  "cfg_scale": 1,
  "cfg_rescale": 0,
  "input_image": "https://api.replicate.com/v1/files/NWE4MWNmNGEtY2EzZi00YTQwLWE5MjctYmIzNjM1NTA2NmYx",
  "use_teacache": true,
  "negative_prompt": "",
  "latent_window_size": 9,
  "distilled_cfg_scale": 10,
  "total_video_length_seconds": 3
}

Output Results

https://replicate.delivery/yhqm/GQqu5Z3vHyI2GtQpvdw0y2EeVUVP3ecT4yxcQFinKRT8unsUA/250514_164835_843_6020_video.mp4

Quick Actions

Use NowView Documentation

Technical Specifications

Hardware Type: A100 (80GB)
Run Count: 1.8k
Commercial Use: Supported
Platform: Replicate

Related Keywords

Progressive video generationAnimating still imagesText-guided motion controlEfficient GPU memorySmooth video blendingCreating dynamic video clipsPrototyping visual narratives

Related Models

Luma Reframe Video

Change the aspect ratio of any video up to 30 seconds long, outputs will be 720p

Frames to Video Merger

Convert a set of image frames (JPG or PNG) into a high-quality MP4 video. Automatically handles sorting and frame order for smooth playback.

twha Video Generation Model

A model for generating videos from text prompts and starting images.