No project description provided
Project description
FFMPerative
Devilishly Simple Video Processing
Large Language Models (LLMs) with Tools can perform complex tasks from natural language prompts. Based on HuggingFace's Agents & Tools, our agent is equipped with a suite of tools for common video processing workflows like:
- Get Video Metadata
- Extract Frame at Frame Number
- Make a Video from a Directory of Images
- Horizontal/Vertical Flip
- Crop Video to Bounding Box
- Speed Up Video by X
- Compress a GIF/Video
- Resize or convert a GIF/Video
- Adjust Audio Levels
Install
Ensure you have ffmpeg installed. On Debian, you can use:
sudo apt-get install ffmpeg
PyPI
Install from pypi with:
pip install ffmperative
Install from source
Clone this repo and install using pip
git clone https://github.com/remyxai/FFMPerative.git
cd FFMPerative/
pip install .
Quickstart
Getting started is easy, import the library and call the ffmp function.
from ffmperative import ffmp
ffmp(prompt="crop video '/path/to/video.mp4' to 200,200,400,400 before writing to '/path/to/video_cropped.mp4', then double the speed of that video and write to '/path/to/video_cropped_fast.mp4'")
CLI
You can also call FFMPerative from the command line, try:
ffmp do --prompt="sample the 5th frame from /path/to/video.mp4"
Roadmap
- Basic Video Tools
- Release to PyPI after Agents are Added to Transformers
- Release LLM checkpoint fine-tuned to use ffmp Tools
Contributing
- We'll gladly review pull requests aimed at improving the library of simple image and video processing tools.
- Interested in contributing to data/templates for specializing an LLM for video processing workflows, ping us!
Resources:
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
No source distribution files available for this release.See tutorial on generating distribution archives.
Built Distribution
Close
Hashes for ffmperative-0.0.2.post1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 3bbcab5dd035a18aed842c3e459a5e0b464e8cf3218e50000bdb2d8080e852cb |
|
MD5 | c395cc4b71965da2e6da498fb94fde74 |
|
BLAKE2b-256 | 7af8e78db7a7b7d03203384894d1618076b304e5ee09cb0d0fca5442ff5c5105 |