mosaico 0.3.0

Open-source video generation framework

Mosaico

Mosaico is a Python library for programmatically creating and managing video compositions. It provides a high-level interface for working with media assets, positioning elements, applying effects, and generating video scripts.

Installation

pip install mosaico

For additional dependencies, see the additional dependencies section in the documentation.

Features

• AI-powered script generation for videos
• Rich media asset management (audio, images, text, subtitles)
• Flexible positioning system (absolute, relative, region-based)
• Built-in effects (pan, zoom) with extensible effect system
• Text-to-speech synthesis integration
• Integration with popular ML frameworks, such as Haystack and LangChain

Quick Start

Install Mosaico and additional dependencies for news video generation:

pip install "mosaico[assemblyai,elevenlabs]"

Easily create and render a video project from a script generator:

import os

from mosaico.audio_transcribers.assemblyai import AssemblyAIAudioTranscriber
from mosaico.script_generators.news import NewsVideoScriptGenerator
from mosaico.speech_synthesizers.elevenlabs import ElevenLabsSpeechSynthesizer
from mosaico.video.project import VideoProject
from mosaico.video.rendering import render_video


# Set your API keys
ANTHROPIC_API_KEY = os.getenv("ANTHROPIC_API_KEY")
ASSEMBLYAI_API_KEY = os.getenv("ASSEMBLYAI_API_KEY")
ELEVENLABS_API_KEY = os.getenv("ELEVENLABS_API_KEY")


# Import your media
media = [
    Media.from_path("background.jpg", metadata={"description": "Background image"}),
    Media.from_path("image1.jpg", metadata={"description": "Image 1"}),
    Media.from_path("image2.jpg", metadata={"description": "Image 2"}),
    Media.from_path("image3.jpg", metadata={"description": "Image 3"}),
]

# Textual context for the video
context = "..."

# Create script generator
script_generator = NewsVideoScriptGenerator(
    context=context,
    language="pt",
    num_paragraphs=8,
    api_key=ANTHROPIC_API_KEY,
)

# Create speech synthesizer
speech_synthesizer = ElevenLabsSpeechSynthesizer(
    voice_id="Xb7hH8MSUJpSbSDYk0k2",
    voice_stability=0.8,
    voice_similarity_boost=0.75,
    voice_speaker_boost=False,
    api_key=ELEVENLABS_API_KEY,
)

# Create audio transcriber for captions
audio_transcriber = AssemblyAIAudioTranscriber(api_key=ASSEMBLYAI_API_KEY)

# Create project
project = (
    VideoProject.from_script_generator(script_generator, media)
    .with_title("My Breaking News Video")
    .with_fps(30)
    .with_resolution((1920, 1080))
    .add_narration(speech_synthesizer)
    .add_captions_from_transcriber(audio_transcriber, overwrite=True)
)

# Render project
render_video(project, "path/to/dir")

Or create a video project from scratch:

from mosaico.video.project import VideoProject
from mosaico.assets import ImageAsset, TextAsset, AudioAsset, AssetReference

# Import your media as production-ready assets
assets = [
    ImageAsset.from_path("background.jpg", metadata={"description": "Background image"}),
    ImageAsset.from_path("image1.jpg", metadata={"description": "Image 1"}),
    ImageAsset.from_path("image2.jpg", metadata={"description": "Image 2"}),
    ImageAsset.from_path("image3.jpg", metadata={"description": "Image 3"}),
    TextAsset.from_data("Subtitle 1"),
    TextAsset.from_data("Subtitle 2"),
    TextAsset.from_data("Subtitle 3"),
    AudioAsset.from_path("narration.mp3"),
]

asset_references = [
    AssetReference.from_asset(background, start_time=0, end_time=10),
    AssetReference.from_asset(image1, start_time=10, end_time=20),
    AssetReference.from_asset(image2, start_time=20, end_time=30),
    AssetReference.from_asset(image3, start_time=30, end_time=40),
    AssetReference.from_asset(subtitle1, start_time=40, end_time=50),
    AssetReference.from_asset(subtitle2, start_time=50, end_time=60),
    AssetReference.from_asset(subtitle3, start_time=60, end_time=70),
    AssetReference.from_asset(narration, start_time=70, end_time=80),
]

scene = Scene(description="My Scene").add_asset_references(asset_references)

project = (
    VideoProject()
    .with_title("My Breaking News Video")
    .with_fps(30)
    .with_resolution((1920, 1080))
    .add_assets(assets)
    # Add the asset references as scene events to the timeline
    .add_timeline_events(scene)
    # Or add asset references directly to the timeline
    # .add_timeline_events(asset_references)
)

# Render project
render_video(project, "path/to/dir")

Cookbook

For common usage patterns and examples, see our Cookbook. Some examples include:

• Creating basic videos with background and text
• Building photo slideshows with music
• Generating news videos from articles
• Working with different asset types
• Applying effects and animations
• Using AI for script generation

Documentation

Comprehensive documentation is available here. Documentation includes:

• Getting Started: Installation, setup, and basic usage
• Concepts: Overview of key concepts and terminology
• Cookbook: Examples and tutorials for common tasks
• API Reference: Detailed reference for all classes and functions
• Development: Information for contributors and developers
• Roadmap: Future plans and features

References

• MoviePy