video-mosaic

Extract frames from a video and arrange them into a single mosaic image.

What it does

video-mosaic takes a video file (or URL), extracts frames, and composes them into a grid — like a contact sheet or storyboard. It includes a header with video metadata and optional timestamps on each frame. Supports YouTube, Twitter/X, TikTok, Instagram, and any site supported by yt-dlp.

┌─────────────────────────────────────────┐
│  filename.mp4                           │
│  1920×1080  •  30.2s  •  30.00 fps      │
├──────┬──────┬──────┬──────┬──────┬──────┤
│  #1  │  #2  │  #3  │  #4  │  #5  │  #6  │
│00:00 │00:05 │00:10 │00:15 │00:20 │00:25 │
├──────┼──────┼──────┼──────┼──────┼──────┤
│  #7  │  #8  │  #9  │ #10  │ #11  │ #12  │
│00:30 │00:35 │00:40 │00:45 │00:50 │00:55 │
└──────┴──────┴──────┴──────┴──────┴──────┘

Installation

With Homebrew (macOS)

brew tap GonzaloFuentes28/tap
brew install video-mosaic

With pip

pip install video-mosaic

From source

git clone https://github.com/GonzaloFuentes28/video-mosaic.git
cd video-mosaic
pip install .

Prerequisites

ffmpeg must be installed and available in your $PATH.

# macOS
brew install ffmpeg

# Ubuntu / Debian
sudo apt install ffmpeg

# Arch
sudo pacman -S ffmpeg

To download videos from URLs (YouTube, Twitter/X, TikTok, etc.), you also need yt-dlp:

brew install yt-dlp   # macOS
pip install yt-dlp     # or via pip

Quick start

# 48 evenly-spaced frames with timestamps
video-mosaic video.mp4 --frames 48 --labels

# One frame every 5 seconds
video-mosaic video.mp4 --every 5s -o contact-sheet.jpg

# Every single frame of a short clip
video-mosaic clip.mp4 --all -o all-frames.png

# One frame per scene change (cuts/transitions)
video-mosaic video.mp4 --scenes -o scenes.jpg

# From a URL (requires yt-dlp)
video-mosaic "https://youtube.com/watch?v=abc" --frames 48 --quality 720
video-mosaic "https://x.com/user/status/123" --frames 24

# Batch mode — process multiple videos at once
video-mosaic *.mp4 --frames 24 -o output/

Usage

video-mosaic INPUT [INPUT ...] [options]

Frame selection (pick one)

Scene detection options

# Detect scene changes with higher sensitivity
video-mosaic video.mp4 --scenes --scene-threshold 0.1 --labels

Trimming

# Frames only from the first minute
video-mosaic video.mp4 --frames 24 --from 0:00 --to 1:00

Filtering

# Clean up a screencast with lots of static frames
video-mosaic screencast.mp4 --every 1s --skip-dupes --skip-black

Layout

# Custom layout: 10 columns, smaller thumbnails, white background
video-mosaic video.mp4 --frames 100 --cols 10 --thumb-width 200 --bg "#ffffff"

Output

# Image + PDF storyboard
video-mosaic video.mp4 --frames 48 --labels --pdf storyboard.pdf

URL downloads

Requires yt-dlp: pip install yt-dlp or brew install yt-dlp.

Batch mode

Pass multiple files to process them all in one command. Outputs are auto-named as {filename}_mosaic.jpg.

# Process all MP4s in the current directory
video-mosaic *.mp4 --frames 24

# Save all mosaics to a specific directory
video-mosaic clip1.mp4 clip2.mp4 clip3.mp4 --frames 24 -o output/

# Batch with PDF export
video-mosaic *.mp4 --scenes --pdf output/

In batch mode, failures on individual videos don't stop the rest — a summary is shown at the end.

Other

Examples

# Quick overview of a movie trailer
video-mosaic trailer.mp4 -o overview.jpg --frames 36 --cols 6 --labels

# Frame-by-frame analysis of a 2-second clip
video-mosaic clip.mp4 -o analysis.png --all --labels

# Every 10 seconds of a long lecture, skip static slides
video-mosaic lecture.mp4 -o sheet.webp --every 10s --skip-dupes --thumb-width 400

# Just the intro of a video, reversed, as PDF
video-mosaic video.mp4 --frames 20 --from 0:00 --to 0:30 --reverse --pdf intro.pdf

# Scene-based mosaic of a music video
video-mosaic "https://youtube.com/watch?v=abc" --scenes --labels --quality 720

Project structure

src/video_mosaic/
├── __init__.py   # Package root, version, VideoMosaicError
├── cli.py        # CLI argument parsing and orchestration
├── download.py   # URL video downloading via yt-dlp
├── probe.py      # Video metadata via ffprobe
├── extract.py    # Frame extraction strategies
├── filters.py    # Black frame / duplicate detection
├── mosaic.py     # Grid composition
├── pdf.py        # PDF storyboard export
└── utils.py      # Timestamp parsing, font loading

Contributing

git clone https://github.com/GonzaloFuentes28/video-mosaic.git
cd video-mosaic
pip install -e ".[dev]"

License

MIT — see LICENSE.